Pipeline run

{
  "final_skills": [
    {
      "is_primary": true,
      "skill_name": "Machine Learning"
    },
    {
      "is_primary": true,
      "skill_name": "Python"
    },
    {
      "is_primary": true,
      "skill_name": "PyTorch"
    },
    {
      "is_primary": true,
      "skill_name": "TensorFlow"
    },
    {
      "is_primary": true,
      "skill_name": "pandas"
    },
    {
      "is_primary": true,
      "skill_name": "NumPy"
    },
    {
      "is_primary": true,
      "skill_name": "OCR"
    },
    {
      "is_primary": true,
      "skill_name": "A/B Testing"
    },
    {
      "is_primary": true,
      "skill_name": "Feature Stores"
    },
    {
      "is_primary": true,
      "skill_name": "Model Serving"
    },
    {
      "is_primary": false,
      "skill_name": "Reinforcement Learning"
    },
    {
      "is_primary": false,
      "skill_name": "Graph Neural Networks"
    },
    {
      "is_primary": false,
      "skill_name": "RAG"
    },
    {
      "is_primary": false,
      "skill_name": "Fine-tuning"
    },
    {
      "is_primary": false,
      "skill_name": "Prompt Engineering"
    },
    {
      "is_primary": true,
      "skill_name": "Classification"
    },
    {
      "is_primary": true,
      "skill_name": "Ranking"
    },
    {
      "is_primary": true,
      "skill_name": "Fraud Detection"
    },
    {
      "is_primary": true,
      "skill_name": "Risk Modeling"
    }
  ],
  "jd_role": {
    "display_name": "Machine Learning Engineer",
    "rationale": null,
    "role_archetype": "Data",
    "slug": ""
  },
  "nano_parsed": {
    "JD_type": "pass",
    "about_company": {
      "source_marker": {
        "first_5_words": "SabPaisa (SRS Live Technologies) is",
        "last_5_words": "to the client\u2019s unique needs."
      },
      "text": "SabPaisa (SRS Live Technologies) is an RBI Authorised Payment Aggregator.\n\nFounded in 2016 with headquarters in New Delhi, a corporate office in Kolkata, and regional offices across the country, it is a rapidly advancing fintech company. SabPaisa is dedicated to providing simplified payment solutions, offering customizable options tailored to the client\u2019s unique needs.",
      "word_count": 64
    },
    "certifications": [],
    "company_name": "SabPaisa (SRS Live Technologies)",
    "ctc": null,
    "domain": {
      "primary": {
        "aliases": [
          "FinTech",
          "Payments"
        ],
        "domain": "Financial Services"
      },
      "secondary": null
    },
    "education": [],
    "experience": {
      "max": 5,
      "min": 3,
      "raw": "3-5 years of applied ML with models deployed to production (not just notebooks/Kaggle)"
    },
    "job_locations": [
      {
        "aliases": [
          "Delhi"
        ],
        "city": "New Delhi",
        "country": "India",
        "state": "Delhi",
        "work_mode": "null"
      },
      {
        "aliases": [
          "Calcutta"
        ],
        "city": "Kolkata",
        "country": "India",
        "state": "West Bengal",
        "work_mode": "null"
      }
    ],
    "role": "Machine Learning Engineer",
    "role_archetype": "Data",
    "roles_and_responsibilities": [
      {
        "bullet_count": 8,
        "heading": "Roles \u0026 Responsibilities",
        "heading_was_present": true,
        "source_marker": {
          "first_5_words": "Build smart payment routing models",
          "last_5_words": "architecture, training procedures, and performance metrics"
        },
        "text": "Build smart payment routing models to optimize transaction success rates across UPI, cards, and net banking rails \nDesign and deploy real-time fraud detection systems with sub-100ms latency\u2014device fingerprinting, behavioral analysis, velocity checks \nDevelop merchant risk scoring models for automated underwriting using documents, transaction patterns, and external signals \nBuild document intelligence pipelines\u2014OCR, classification, and data extraction for KYC documents (PAN, Aadhaar, GST, bank statements) \nSet up ML infrastructure\u2014feature stores, model serving, A/B testing frameworks, monitoring and alerting \nCollaborate with product and engineering teams to integrate ML models into production systems \nMonitor model performance, detect drift, and implement retraining pipelines \nDocument model architecture, training procedures, and performance metrics",
        "word_count": 134
      },
      {
        "bullet_count": 5,
        "heading": "Requirements",
        "heading_was_present": true,
        "source_marker": {
          "first_5_words": "3-5 years of applied ML",
          "last_5_words": "feature stores, model serving"
        },
        "text": "3-5 years of applied ML with models deployed to production (not just notebooks/Kaggle) \nStrong Python with PyTorch or TensorFlow, plus pandas/numpy for data manipulation \nEnd-to-end ML skills: feature engineering, training, evaluation, deployment, monitoring \nExperience with tabular/transactional data and classification/ranking problems \nUnderstanding of real-time inference\u2014latency budgets, feature stores, model serving",
        "word_count": 66
      },
      {
        "bullet_count": 4,
        "heading": "Good to Have",
        "heading_was_present": true,
        "source_marker": {
          "first_5_words": "Fraud detection or risk modeling",
          "last_5_words": "fine-tuning, prompt engineering"
        },
        "text": "Fraud detection or risk modeling experience in fintech/payments \nMulti-armed bandits or reinforcement learning for optimization \nGraph neural networks for network-based detection \nLLM experience: RAG pipelines, fine-tuning, prompt engineering",
        "word_count": 40
      }
    ],
    "urls": [
      {
        "type": "other",
        "url": "https://www.youtube.com/watch?v=K7Z7A059faE"
      }
    ]
  },
  "run_id": null
}

{
  "alias_matches": [
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 3647,
      "existing_alias_text": "Machine Learning",
      "input_term": "Machine Learning",
      "matched_canonical": {
        "category_id": 2,
        "display_name": "Machine Learning",
        "id": 2672,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "CONCEPT",
        "slug": "machine-learning",
        "sub_category_id": 2146,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 608,
      "existing_alias_text": "Python",
      "input_term": "Python",
      "matched_canonical": {
        "category_id": 5,
        "display_name": "Python",
        "id": 393,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LANGUAGE",
        "slug": "python",
        "sub_category_id": 54,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 861,
      "existing_alias_text": "PyTorch",
      "input_term": "PyTorch",
      "matched_canonical": {
        "category_id": 6,
        "display_name": "PyTorch",
        "id": 557,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "pytorch",
        "sub_category_id": 456,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 862,
      "existing_alias_text": "TensorFlow",
      "input_term": "TensorFlow",
      "matched_canonical": {
        "category_id": 6,
        "display_name": "TensorFlow",
        "id": 558,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "tensorflow",
        "sub_category_id": 456,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 766,
      "existing_alias_text": "pandas",
      "input_term": "pandas",
      "matched_canonical": {
        "category_id": 6,
        "display_name": "pandas",
        "id": 474,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "pandas",
        "sub_category_id": 454,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 767,
      "existing_alias_text": "NumPy",
      "input_term": "NumPy",
      "matched_canonical": {
        "category_id": 6,
        "display_name": "NumPy",
        "id": 475,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "numpy",
        "sub_category_id": 459,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 3610,
      "existing_alias_text": "OCR",
      "input_term": "OCR",
      "matched_canonical": {
        "category_id": 2,
        "display_name": "OCR",
        "id": 2653,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "CONCEPT",
        "slug": "ocr",
        "sub_category_id": 2166,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 798,
      "existing_alias_text": "A/B testing",
      "input_term": "A/B Testing",
      "matched_canonical": {
        "category_id": 7,
        "display_name": "A/B testing",
        "id": 506,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "METHODOLOGY",
        "slug": "a-b-testing",
        "sub_category_id": 469,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 3616,
      "existing_alias_text": "RAG",
      "input_term": "RAG",
      "matched_canonical": {
        "category_id": 2,
        "display_name": "RAG",
        "id": 2659,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "CONCEPT",
        "slug": "rag",
        "sub_category_id": 2172,
        "typical_lifespan": "EVERGREEN",
        "volatility": "EMERGING"
      },
      "matched_via": "alias"
    },
    {
      "alias_persist_skipped_reason": "alias_text already exists for this canonical skill",
      "alias_persisted": false,
      "existing_alias_id": 3658,
      "existing_alias_text": "Prompt Engineering",
      "input_term": "Prompt Engineering",
      "matched_canonical": {
        "category_id": 7,
        "display_name": "Prompt Engineering",
        "id": 2683,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "METHODOLOGY",
        "slug": "prompt-engineering",
        "sub_category_id": 2191,
        "typical_lifespan": "EVERGREEN",
        "volatility": "EMERGING"
      },
      "matched_via": "alias"
    }
  ],
  "candidate_roles": [
    {
      "display_name": "Data Analyst",
      "id": 20,
      "rationale": null,
      "role_archetype": null,
      "slug": "data-analyst",
      "source": "db"
    },
    {
      "display_name": "Data Scientist",
      "id": 7,
      "rationale": null,
      "role_archetype": null,
      "slug": "data-scientist",
      "source": "db"
    },
    {
      "display_name": "Azure Cloud Engineer",
      "id": 4,
      "rationale": null,
      "role_archetype": null,
      "slug": "azure-cloud-engineer",
      "source": "db"
    },
    {
      "display_name": "Cloud Engineer",
      "id": 18,
      "rationale": null,
      "role_archetype": null,
      "slug": "cloud-engineer",
      "source": "db"
    },
    {
      "display_name": "Virtualization Engineer",
      "id": 26,
      "rationale": null,
      "role_archetype": null,
      "slug": "virtualization-engineer",
      "source": "db"
    },
    {
      "display_name": "Network Engineer",
      "id": 21,
      "rationale": null,
      "role_archetype": null,
      "slug": "network-engineer",
      "source": "db"
    },
    {
      "display_name": "AI Engineer",
      "id": 12,
      "rationale": null,
      "role_archetype": null,
      "slug": "ai-engineer",
      "source": "db"
    },
    {
      "display_name": "Backend Engineer",
      "id": 14,
      "rationale": null,
      "role_archetype": null,
      "slug": "backend-engineer",
      "source": "db"
    },
    {
      "display_name": "Data Engineer",
      "id": 6,
      "rationale": null,
      "role_archetype": null,
      "slug": "data-engineer",
      "source": "db"
    },
    {
      "display_name": "Machine Learning Engineer",
      "id": 10,
      "rationale": null,
      "role_archetype": null,
      "slug": "machine-learning-engineer",
      "source": "db"
    },
    {
      "display_name": "Cybersecurity Engineer",
      "id": 9,
      "rationale": null,
      "role_archetype": null,
      "slug": "cybersecurity-engineer",
      "source": "db"
    },
    {
      "display_name": "Automation Tester",
      "id": 16,
      "rationale": null,
      "role_archetype": null,
      "slug": "automation-tester",
      "source": "db"
    },
    {
      "display_name": "Cloud Architect",
      "id": 11,
      "rationale": null,
      "role_archetype": null,
      "slug": "cloud-architect",
      "source": "db"
    },
    {
      "display_name": "MLOps Engineer",
      "id": 5,
      "rationale": null,
      "role_archetype": null,
      "slug": "mlops-engineer",
      "source": "db"
    }
  ],
  "chosen_role": {
    "display_name": "Machine Learning Engineer",
    "id": 10,
    "rationale": "The primary skills include Machine Learning, Python, and various deep learning frameworks, aligning well with the role of a Machine Learning Engineer.",
    "role_archetype": null,
    "slug": "machine-learning-engineer",
    "source": "db"
  },
  "dimensions": [
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Project Delivery and Coordination",
        "id": 366,
        "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
        "slug": "d_init_02",
        "source": "db"
      },
      "input_skill": "Machine Learning",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Machine Learning",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Analytical Programming Languages",
        "id": 82,
        "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
        "slug": "analytical-programming-languages",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Analyst",
          "id": 20,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-analyst",
          "source": "db"
        },
        {
          "display_name": "Data Scientist",
          "id": 7,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-scientist",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Automation Scripting and CLI",
        "id": 48,
        "rationale": "Uses scripts and command-line tooling to execute repeatable Azure operations and reduce manual work. This is a practical cluster because the role frequently automates provisioning, checks, and remediation tasks.",
        "slug": "automation-scripting-and-cli",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Azure Cloud Engineer",
          "id": 4,
          "rationale": null,
          "role_archetype": null,
          "slug": "azure-cloud-engineer",
          "source": "db"
        },
        {
          "display_name": "Cloud Engineer",
          "id": 18,
          "rationale": null,
          "role_archetype": null,
          "slug": "cloud-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Automation and Scripting for Operations",
        "id": 361,
        "rationale": "Scripts and lightweight automation used to execute repetitive virtualization tasks and enforce operational consistency. This is the practical glue that reduces manual host and VM administration.",
        "slug": "automation-and-scripting-for-operations",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Virtualization Engineer",
          "id": 26,
          "rationale": null,
          "role_archetype": null,
          "slug": "virtualization-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Network Automation and Scripting",
        "id": 285,
        "rationale": "Covers scripts and automation used to configure, validate, and audit network devices and services. This cluster is coherent because repeatable network operations increasingly depend on programmatic changes and checks.",
        "slug": "network-automation-and-scripting",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Network Engineer",
          "id": 21,
          "rationale": null,
          "role_archetype": null,
          "slug": "network-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Programming Languages for AI Workflows",
        "id": 261,
        "rationale": "Languages used to implement AI feature logic, orchestration, and response handling inside product code. This is the core coding surface for turning prompts and model calls into reliable application behavior.",
        "slug": "programming-languages-for-ai-workflows",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "AI Engineer",
          "id": 12,
          "rationale": null,
          "role_archetype": null,
          "slug": "ai-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Programming Languages for Backend Systems",
        "id": 140,
        "rationale": "Languages used to implement server-side business logic, request handlers, workers, and service integrations. This is the core coding surface for backend feature delivery and maintenance.",
        "slug": "programming-languages-for-backend-systems",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Backend Engineer",
          "id": 14,
          "rationale": null,
          "role_archetype": null,
          "slug": "backend-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Programming Languages for Data Work",
        "id": 67,
        "rationale": "Languages used to implement data pipelines, transformations, and operational utilities. This is the code layer for expressing extraction, parsing, validation, and orchestration logic in data engineering workflows.",
        "slug": "programming-languages-for-data-work",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Engineer",
          "id": 6,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Programming Languages for ML Systems",
        "id": 113,
        "rationale": "Languages used to implement model integration code, inference services, and feature-processing logic. This is the core coding surface for turning trained models into product-facing software components.",
        "slug": "programming-languages-for-ml-systems",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Machine Learning Engineer",
          "id": 10,
          "rationale": null,
          "role_archetype": null,
          "slug": "machine-learning-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Programming Languages for Security Work",
        "id": 328,
        "rationale": "Languages used to automate security tasks, write detection logic, and build analysis or remediation tooling. This is the core coding surface for a cybersecurity engineer across scripts, queries, and small utilities.",
        "slug": "programming-languages-for-security-work",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Cybersecurity Engineer",
          "id": 9,
          "rationale": null,
          "role_archetype": null,
          "slug": "cybersecurity-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Programming Languages for Test Automation",
        "id": 193,
        "rationale": "Languages used to implement automated checks, helper utilities, and test harness code. This is the core coding surface for turning test ideas into maintainable automation.",
        "slug": "programming-languages-for-test-automation",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Automation Tester",
          "id": 16,
          "rationale": null,
          "role_archetype": null,
          "slug": "automation-tester",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Security Automation and Scripting",
        "id": 258,
        "rationale": "Automating repeatable security checks, enrichment, and remediation workflows. This cluster is coherent because the role often needs lightweight automation to scale analysis and response.",
        "slug": "security-automation-and-scripting",
        "source": "db"
      },
      "input_skill": "Python",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Cybersecurity Engineer",
          "id": 9,
          "rationale": null,
          "role_archetype": null,
          "slug": "cybersecurity-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Applied Machine Learning Toolkits",
        "id": 94,
        "rationale": "Libraries and frameworks used to prototype and compare models quickly. This dimension captures the concrete tooling layer beneath modeling methods and evaluation.",
        "slug": "applied-machine-learning-toolkits",
        "source": "db"
      },
      "input_skill": "PyTorch",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Scientist",
          "id": 7,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-scientist",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Applied Machine Learning Toolkits",
        "id": 94,
        "rationale": "Libraries and frameworks used to prototype and compare models quickly. This dimension captures the concrete tooling layer beneath modeling methods and evaluation.",
        "slug": "applied-machine-learning-toolkits",
        "source": "db"
      },
      "input_skill": "TensorFlow",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Scientist",
          "id": 7,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-scientist",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Analytical Programming Languages",
        "id": 82,
        "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
        "slug": "analytical-programming-languages",
        "source": "db"
      },
      "input_skill": "pandas",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Analyst",
          "id": 20,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-analyst",
          "source": "db"
        },
        {
          "display_name": "Data Scientist",
          "id": 7,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-scientist",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Analytical Programming Languages",
        "id": 82,
        "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
        "slug": "analytical-programming-languages",
        "source": "db"
      },
      "input_skill": "NumPy",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Analyst",
          "id": 20,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-analyst",
          "source": "db"
        },
        {
          "display_name": "Data Scientist",
          "id": 7,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-scientist",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Project Delivery and Coordination",
        "id": 366,
        "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
        "slug": "d_init_02",
        "source": "db"
      },
      "input_skill": "OCR",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "OCR",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Experiment Design and Analysis",
        "id": 87,
        "rationale": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This cluster is central when the role needs to compare interventions, treatments, or product changes with rigor.",
        "slug": "experiment-design-and-analysis",
        "source": "db"
      },
      "input_skill": "A/B Testing",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Data Scientist",
          "id": 7,
          "rationale": null,
          "role_archetype": null,
          "slug": "data-scientist",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Cloud Service Integration Patterns",
        "id": 188,
        "rationale": "Covers how cloud services and workloads connect through APIs, events, shared services, and integration boundaries. This cluster is coherent because architects must define interaction patterns that preserve decoupling, security, and operability.",
        "slug": "cloud-service-integration-patterns",
        "source": "db"
      },
      "input_skill": "RAG",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "Cloud Architect",
          "id": 11,
          "rationale": null,
          "role_archetype": null,
          "slug": "cloud-architect",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Context Management and Retrieval",
        "id": 264,
        "rationale": "Preparing, selecting, and packaging context for model calls so responses stay relevant and grounded. This is a distinct cluster because AI features often depend on what information is included, summarized, or retrieved at call time.",
        "slug": "context-management-and-retrieval",
        "source": "db"
      },
      "input_skill": "RAG",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "AI Engineer",
          "id": 12,
          "rationale": null,
          "role_archetype": null,
          "slug": "ai-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Context Management and Retrieval",
        "id": 264,
        "rationale": "Preparing, selecting, and packaging context for model calls so responses stay relevant and grounded. This is a distinct cluster because AI features often depend on what information is included, summarized, or retrieved at call time.",
        "slug": "context-management-and-retrieval",
        "source": "db"
      },
      "input_skill": "Prompt Engineering",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "AI Engineer",
          "id": 12,
          "rationale": null,
          "role_archetype": null,
          "slug": "ai-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Feature Stores",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": null,
        "display_name": "Production Model Serving Architecture",
        "id": null,
        "rationale": "Patterns for hosting, routing, versioning, and scaling model inference in production systems, including online inference endpoints, batch scoring services, streaming inference, and serving topologies such as dedicated services, sidecars, gateways, or embedded serving layers. This skill covers request routing, autoscaling serving replicas, canary rollout for models, model version selection, and latency/throughput tradeoffs.",
        "slug": "d_merge_01",
        "source": "llm"
      },
      "input_skill": "Model Serving",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Model Serving Deployment and Runtime Packaging",
        "id": 52,
        "rationale": "Operational deployment of trained models into online, batch, or streaming serving environments, including packaging models and model servers into containers or managed inference runtimes, coordinating rollout, and handing off to inference systems. Covers serving frameworks and platforms such as TensorFlow Serving, TorchServe, Triton Inference Server, BentoML, KServe, and Seldon Core, plus container/runtime concerns like Docker images, GPU-enabled containers, base image selection, container entrypoints, runtime dependencies, and image scanning for model services.",
        "slug": "model-serving-deployment-and-runtime-packaging",
        "source": "db"
      },
      "input_skill": "Model Serving",
      "llm_role": null,
      "roles_from_db": [
        {
          "display_name": "MLOps Engineer",
          "id": 5,
          "rationale": null,
          "role_archetype": null,
          "slug": "mlops-engineer",
          "source": "db"
        },
        {
          "display_name": "Machine Learning Engineer",
          "id": 10,
          "rationale": null,
          "role_archetype": null,
          "slug": "machine-learning-engineer",
          "source": "db"
        }
      ]
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Reinforcement Learning",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Graph Neural Networks",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Project Delivery and Coordination",
        "id": 366,
        "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
        "slug": "d_init_02",
        "source": "db"
      },
      "input_skill": "Graph Neural Networks",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": null,
        "display_name": "Experimental Design, Testing, and Causal Analysis",
        "id": null,
        "rationale": "Designing controlled experiments and analyzing results to estimate causal impact. This includes A/B tests, randomized experiments, hypothesis testing, metric selection, statistical significance, treatment/control design, power analysis, causal inference, and evaluating model or product variants through rigorous comparison. It covers the experimental method and interpretation of outcomes, not the mechanics of model training or data preparation.",
        "slug": "d_merge_01",
        "source": "llm"
      },
      "input_skill": "Fine-tuning",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Fine-tuning",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Classification",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Ranking",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Fraud Detection",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Project Delivery and Coordination",
        "id": 366,
        "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
        "slug": "d_init_02",
        "source": "db"
      },
      "input_skill": "Fraud Detection",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": null,
        "display_name": "Experiment Design, Causal Inference, and Analysis",
        "id": null,
        "rationale": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This includes A/B testing, randomized controlled trials, hypothesis testing, confidence intervals, causal inference, uplift analysis, backtesting, and treatment effect estimation. The focus is on rigorous causal measurement of interventions, treatments, or product changes, not on risk modeling, deployment, or reporting.",
        "slug": "d_merge_01",
        "source": "llm"
      },
      "input_skill": "Risk Modeling",
      "llm_role": null,
      "roles_from_db": []
    },
    {
      "dimension": {
        "difficulty_hint": "well_known",
        "display_name": "Version Control Systems",
        "id": 365,
        "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
        "slug": "d_init_01",
        "source": "db"
      },
      "input_skill": "Risk Modeling",
      "llm_role": null,
      "roles_from_db": []
    }
  ],
  "input_final_skills": [
    "Machine Learning",
    "Python",
    "PyTorch",
    "TensorFlow",
    "pandas",
    "NumPy",
    "OCR",
    "A/B Testing",
    "Feature Stores",
    "Model Serving",
    "Reinforcement Learning",
    "Graph Neural Networks",
    "RAG",
    "Fine-tuning",
    "Prompt Engineering",
    "Classification",
    "Ranking",
    "Fraud Detection",
    "Risk Modeling"
  ],
  "input_llm_skills": [
    "Machine Learning",
    "Python",
    "PyTorch",
    "TensorFlow",
    "pandas",
    "NumPy",
    "OCR",
    "A/B Testing",
    "Feature Stores",
    "Model Serving",
    "Reinforcement Learning",
    "Graph Neural Networks",
    "RAG",
    "Fine-tuning",
    "Prompt Engineering",
    "Classification",
    "Ranking",
    "Fraud Detection",
    "Risk Modeling"
  ],
  "new_aliases_persisted": 0,
  "run_id": "317b22e6-ddd9-4256-bb2b-552034ba2d8e",
  "skills_detail": [
    {
      "aliases_in_db": [
        {
          "alias_text": "Machine Learning",
          "alias_type": "CANONICAL",
          "id": 3647,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 2,
        "display_name": "Machine Learning",
        "id": 2672,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "CONCEPT",
        "slug": "machine-learning",
        "sub_category_id": 2146,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Project Delivery and Coordination",
            "id": 366,
            "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
            "slug": "d_init_02",
            "source": "db"
          },
          "input_skill": "Machine Learning",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Machine Learning",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Machine Learning",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "Python",
          "alias_type": "CANONICAL",
          "id": 608,
          "is_primary": true,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 2",
          "alias_type": "VERSION",
          "id": 611,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 2.x",
          "alias_type": "VERSION",
          "id": 613,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 3",
          "alias_type": "VERSION",
          "id": 612,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 3.10",
          "alias_type": "VERSION",
          "id": 2330,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 3.11",
          "alias_type": "VERSION",
          "id": 2331,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 3.12",
          "alias_type": "VERSION",
          "id": 2332,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "Python 3.x",
          "alias_type": "VERSION",
          "id": 614,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "py2",
          "alias_type": "VERSION",
          "id": 609,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "py3",
          "alias_type": "VERSION",
          "id": 610,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 2",
          "alias_type": "VERSION",
          "id": 2152,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 2.x",
          "alias_type": "VERSION",
          "id": 2154,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 3",
          "alias_type": "VERSION",
          "id": 990,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 3.10",
          "alias_type": "VERSION",
          "id": 992,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 3.11",
          "alias_type": "VERSION",
          "id": 993,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 3.12",
          "alias_type": "VERSION",
          "id": 994,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python 3.x",
          "alias_type": "VERSION",
          "id": 991,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python2",
          "alias_type": "VERSION",
          "id": 2150,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "python3",
          "alias_type": "VERSION",
          "id": 989,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 5,
        "display_name": "Python",
        "id": 393,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LANGUAGE",
        "slug": "python",
        "sub_category_id": 54,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Analytical Programming Languages",
            "id": 82,
            "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
            "slug": "analytical-programming-languages",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Analyst",
              "id": 20,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-analyst",
              "source": "db"
            },
            {
              "display_name": "Data Scientist",
              "id": 7,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-scientist",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Automation Scripting and CLI",
            "id": 48,
            "rationale": "Uses scripts and command-line tooling to execute repeatable Azure operations and reduce manual work. This is a practical cluster because the role frequently automates provisioning, checks, and remediation tasks.",
            "slug": "automation-scripting-and-cli",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Azure Cloud Engineer",
              "id": 4,
              "rationale": null,
              "role_archetype": null,
              "slug": "azure-cloud-engineer",
              "source": "db"
            },
            {
              "display_name": "Cloud Engineer",
              "id": 18,
              "rationale": null,
              "role_archetype": null,
              "slug": "cloud-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Automation and Scripting for Operations",
            "id": 361,
            "rationale": "Scripts and lightweight automation used to execute repetitive virtualization tasks and enforce operational consistency. This is the practical glue that reduces manual host and VM administration.",
            "slug": "automation-and-scripting-for-operations",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Virtualization Engineer",
              "id": 26,
              "rationale": null,
              "role_archetype": null,
              "slug": "virtualization-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Network Automation and Scripting",
            "id": 285,
            "rationale": "Covers scripts and automation used to configure, validate, and audit network devices and services. This cluster is coherent because repeatable network operations increasingly depend on programmatic changes and checks.",
            "slug": "network-automation-and-scripting",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Network Engineer",
              "id": 21,
              "rationale": null,
              "role_archetype": null,
              "slug": "network-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Programming Languages for AI Workflows",
            "id": 261,
            "rationale": "Languages used to implement AI feature logic, orchestration, and response handling inside product code. This is the core coding surface for turning prompts and model calls into reliable application behavior.",
            "slug": "programming-languages-for-ai-workflows",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "AI Engineer",
              "id": 12,
              "rationale": null,
              "role_archetype": null,
              "slug": "ai-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Programming Languages for Backend Systems",
            "id": 140,
            "rationale": "Languages used to implement server-side business logic, request handlers, workers, and service integrations. This is the core coding surface for backend feature delivery and maintenance.",
            "slug": "programming-languages-for-backend-systems",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Backend Engineer",
              "id": 14,
              "rationale": null,
              "role_archetype": null,
              "slug": "backend-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Programming Languages for Data Work",
            "id": 67,
            "rationale": "Languages used to implement data pipelines, transformations, and operational utilities. This is the code layer for expressing extraction, parsing, validation, and orchestration logic in data engineering workflows.",
            "slug": "programming-languages-for-data-work",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Engineer",
              "id": 6,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Programming Languages for ML Systems",
            "id": 113,
            "rationale": "Languages used to implement model integration code, inference services, and feature-processing logic. This is the core coding surface for turning trained models into product-facing software components.",
            "slug": "programming-languages-for-ml-systems",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Machine Learning Engineer",
              "id": 10,
              "rationale": null,
              "role_archetype": null,
              "slug": "machine-learning-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Programming Languages for Security Work",
            "id": 328,
            "rationale": "Languages used to automate security tasks, write detection logic, and build analysis or remediation tooling. This is the core coding surface for a cybersecurity engineer across scripts, queries, and small utilities.",
            "slug": "programming-languages-for-security-work",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Cybersecurity Engineer",
              "id": 9,
              "rationale": null,
              "role_archetype": null,
              "slug": "cybersecurity-engineer",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Programming Languages for Test Automation",
            "id": 193,
            "rationale": "Languages used to implement automated checks, helper utilities, and test harness code. This is the core coding surface for turning test ideas into maintainable automation.",
            "slug": "programming-languages-for-test-automation",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Automation Tester",
              "id": 16,
              "rationale": null,
              "role_archetype": null,
              "slug": "automation-tester",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Security Automation and Scripting",
            "id": 258,
            "rationale": "Automating repeatable security checks, enrichment, and remediation workflows. This cluster is coherent because the role often needs lightweight automation to scale analysis and response.",
            "slug": "security-automation-and-scripting",
            "source": "db"
          },
          "input_skill": "Python",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Cybersecurity Engineer",
              "id": 9,
              "rationale": null,
              "role_archetype": null,
              "slug": "cybersecurity-engineer",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "Python",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "PyTorch",
          "alias_type": "CANONICAL",
          "id": 861,
          "is_primary": true,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 6,
        "display_name": "PyTorch",
        "id": 557,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "pytorch",
        "sub_category_id": 456,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Applied Machine Learning Toolkits",
            "id": 94,
            "rationale": "Libraries and frameworks used to prototype and compare models quickly. This dimension captures the concrete tooling layer beneath modeling methods and evaluation.",
            "slug": "applied-machine-learning-toolkits",
            "source": "db"
          },
          "input_skill": "PyTorch",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Scientist",
              "id": 7,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-scientist",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "PyTorch",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "TensorFlow",
          "alias_type": "CANONICAL",
          "id": 862,
          "is_primary": true,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "TF1",
          "alias_type": "VERSION",
          "id": 863,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "TF2",
          "alias_type": "VERSION",
          "id": 864,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "TensorFlow 1",
          "alias_type": "VERSION",
          "id": 865,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "TensorFlow 1.x",
          "alias_type": "VERSION",
          "id": 867,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "TensorFlow 2",
          "alias_type": "VERSION",
          "id": 866,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "TensorFlow 2.x",
          "alias_type": "VERSION",
          "id": 868,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 6,
        "display_name": "TensorFlow",
        "id": 558,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "tensorflow",
        "sub_category_id": 456,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Applied Machine Learning Toolkits",
            "id": 94,
            "rationale": "Libraries and frameworks used to prototype and compare models quickly. This dimension captures the concrete tooling layer beneath modeling methods and evaluation.",
            "slug": "applied-machine-learning-toolkits",
            "source": "db"
          },
          "input_skill": "TensorFlow",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Scientist",
              "id": 7,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-scientist",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "TensorFlow",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "pandas",
          "alias_type": "CANONICAL",
          "id": 766,
          "is_primary": true,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "pandas 1.x",
          "alias_type": "VERSION",
          "id": 2529,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "pandas 2.x",
          "alias_type": "VERSION",
          "id": 2530,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "pandas1",
          "alias_type": "VERSION",
          "id": 2527,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        },
        {
          "alias_text": "pandas2",
          "alias_type": "VERSION",
          "id": 2528,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 6,
        "display_name": "pandas",
        "id": 474,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "pandas",
        "sub_category_id": 454,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Analytical Programming Languages",
            "id": 82,
            "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
            "slug": "analytical-programming-languages",
            "source": "db"
          },
          "input_skill": "pandas",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Analyst",
              "id": 20,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-analyst",
              "source": "db"
            },
            {
              "display_name": "Data Scientist",
              "id": 7,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-scientist",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "pandas",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "NumPy",
          "alias_type": "CANONICAL",
          "id": 767,
          "is_primary": true,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 6,
        "display_name": "NumPy",
        "id": 475,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "LIBRARY",
        "slug": "numpy",
        "sub_category_id": 459,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Analytical Programming Languages",
            "id": 82,
            "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
            "slug": "analytical-programming-languages",
            "source": "db"
          },
          "input_skill": "NumPy",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Analyst",
              "id": 20,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-analyst",
              "source": "db"
            },
            {
              "display_name": "Data Scientist",
              "id": 7,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-scientist",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "NumPy",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "OCR",
          "alias_type": "CANONICAL",
          "id": 3610,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 2,
        "display_name": "OCR",
        "id": 2653,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "CONCEPT",
        "slug": "ocr",
        "sub_category_id": 2166,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Project Delivery and Coordination",
            "id": 366,
            "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
            "slug": "d_init_02",
            "source": "db"
          },
          "input_skill": "OCR",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "OCR",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "OCR",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "A/B testing",
          "alias_type": "CANONICAL",
          "id": 798,
          "is_primary": true,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 7,
        "display_name": "A/B testing",
        "id": 506,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "METHODOLOGY",
        "slug": "a-b-testing",
        "sub_category_id": 469,
        "typical_lifespan": "EVERGREEN",
        "volatility": "STABLE"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Experiment Design and Analysis",
            "id": 87,
            "rationale": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This cluster is central when the role needs to compare interventions, treatments, or product changes with rigor.",
            "slug": "experiment-design-and-analysis",
            "source": "db"
          },
          "input_skill": "A/B Testing",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Data Scientist",
              "id": 7,
              "rationale": null,
              "role_archetype": null,
              "slug": "data-scientist",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "A/B Testing",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Feature Stores",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Feature Stores",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Datastore",
          "skill_nature": "TOOL",
          "sub_category": "feature_store",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "EMERGING"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "\u201cFeature Stores\u201d is a fairly specific ML infrastructure term and is unlikely to be mistaken for a different catalog skill in typical job descriptions."
          },
          "context_keywords": {
            "context_keywords": [
              "offline store",
              "online store",
              "feature registry",
              "feature engineering",
              "point-in-time joins",
              "training-serving skew",
              "feature pipelines",
              "materialized features",
              "feature views",
              "entity keys",
              "batch ingestion",
              "streaming ingestion",
              "low-latency serving",
              "feature freshness",
              "ML platform"
            ]
          },
          "maturity": {
            "confidence": 0.82,
            "maturity": "emerging",
            "reasoning": "Feature stores appear in more ML platform and MLOps job descriptions, but they\u2019re still far from universal; market demand is growing alongside tools like Feast and managed offerings from major cloud vendors."
          },
          "skill_id": "feature-stores",
          "vendor_license": {
            "confidence": 0.98,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Systems for defining, storing, versioning, and serving reusable ML features for training and online inference. Feature stores belong here because they manage feature computation, consistency, and low-latency retrieval across pipelines and model-serving workflows.",
            "exemplar_skills": [
              "Feature Stores",
              "feature registry",
              "offline feature store",
              "online feature store",
              "point-in-time joins",
              "feature materialization",
              "feature freshness",
              "feature lineage",
              "Feast",
              "Tecton"
            ],
            "in_scope": "Feature Stores, feature definitions, feature registry, offline feature store, online feature store, point-in-time joins, feature versioning, feature materialization, feature serving, feature freshness, feature lineage, Feast, Tecton, Hopsworks",
            "name": "Feature Store Systems",
            "out_of_scope": "Model training algorithms and hyperparameter tuning, model serving infrastructure and request routing, raw data warehouse modeling and SQL optimization, stream processing infrastructure not specific to features",
            "overlap_flags": [
              {
                "reason": "Feature stores often depend on batch and streaming pipelines to compute and refresh features, but the store itself is the reusable feature layer.",
                "with_dim_id": "inference-data-pipelines",
                "with_dim_name": null,
                "with_role": "MLOps Engineer"
              },
              {
                "reason": "Online feature retrieval is frequently used in serving paths, but this dimension focuses on model deployment rather than feature management.",
                "with_dim_id": "model-serving-deployment-and-runtime-packaging",
                "with_dim_name": null,
                "with_role": "MLOps Engineer, Machine Learning Engineer"
              },
              {
                "reason": "Feature stores rely on efficient storage and retrieval patterns, though query tuning is a supporting concern rather than the core concept.",
                "with_dim_id": "data-access-and-query-optimization",
                "with_dim_name": null,
                "with_role": "Data Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          }
        ],
        "merge_log": [],
        "placed": {
          "name": "Feature Stores",
          "placement_confidence": 0.92,
          "primary_dimension": "d_init_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 1 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [],
          "skill_id": "feature-stores"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [
            "data-stores"
          ],
          "related_to": [
            "lakefs",
            "storage-layout",
            "nosql",
            "postgresql",
            "snapshot",
            "data-modeling",
            "data-pipelines",
            "stream-processing",
            "derived-views",
            "subgraphs"
          ],
          "requires": [],
          "skill_id": "feature-stores",
          "suppress_on_match": []
        },
        "skill_id": "feature-stores",
        "split_log": [],
        "typed": {
          "alternatives_considered": [
            "Platform: ruled out \u2014 feature stores are usually a data system component rather than a full hosted multi-tenant environment.",
            "Concept: ruled out \u2014 this is a concrete system category, not just a knowledge unit."
          ],
          "confidence": 0.78,
          "name": "Feature Stores",
          "reasoning": "Feature Stores are primarily systems that persist and serve reusable features, so by the Datastore vs Format rule they fit Datastore rather than a tool or framework.",
          "skill_id": "feature-stores",
          "subtype": "feature_store",
          "type": "Datastore"
        },
        "warnings": []
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": null,
            "display_name": "Production Model Serving Architecture",
            "id": null,
            "rationale": "Patterns for hosting, routing, versioning, and scaling model inference in production systems, including online inference endpoints, batch scoring services, streaming inference, and serving topologies such as dedicated services, sidecars, gateways, or embedded serving layers. This skill covers request routing, autoscaling serving replicas, canary rollout for models, model version selection, and latency/throughput tradeoffs.",
            "slug": "d_merge_01",
            "source": "llm"
          },
          "input_skill": "Model Serving",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Model Serving Deployment and Runtime Packaging",
            "id": 52,
            "rationale": "Operational deployment of trained models into online, batch, or streaming serving environments, including packaging models and model servers into containers or managed inference runtimes, coordinating rollout, and handing off to inference systems. Covers serving frameworks and platforms such as TensorFlow Serving, TorchServe, Triton Inference Server, BentoML, KServe, and Seldon Core, plus container/runtime concerns like Docker images, GPU-enabled containers, base image selection, container entrypoints, runtime dependencies, and image scanning for model services.",
            "slug": "model-serving-deployment-and-runtime-packaging",
            "source": "db"
          },
          "input_skill": "Model Serving",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "MLOps Engineer",
              "id": 5,
              "rationale": null,
              "role_archetype": null,
              "slug": "mlops-engineer",
              "source": "db"
            },
            {
              "display_name": "Machine Learning Engineer",
              "id": 10,
              "rationale": null,
              "role_archetype": null,
              "slug": "machine-learning-engineer",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "Model Serving",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Architecture",
          "skill_nature": "PATTERN",
          "sub_category": "model_serving_architecture",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "EMERGING"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "Model Serving is a fairly specific ML infrastructure term. In typical JDs it is unlikely to be mistaken for a different catalog skill, since the phrase usually clearly refers to deploying and exposing models for inference."
          },
          "context_keywords": {
            "context_keywords": [
              "inference",
              "latency",
              "throughput",
              "autoscaling",
              "canary deployment",
              "A/B testing",
              "gRPC",
              "REST API",
              "Kubernetes",
              "Docker",
              "TensorFlow Serving",
              "TorchServe",
              "NVIDIA Triton",
              "batch inference",
              "model registry"
            ]
          },
          "maturity": {
            "confidence": 0.84,
            "maturity": "emerging",
            "reasoning": "Model serving is increasingly listed in ML platform and MLOps job descriptions, with vendors like AWS SageMaker, Vertex AI, and KServe driving adoption, but it is not yet a universal hiring staple."
          },
          "skill_id": "model-serving",
          "vendor_license": {
            "confidence": 0.98,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [
          {
            "a_dim_id": "model-serving-architecture",
            "a_name": "Model Serving Architecture",
            "a_role": "__skill_focal__",
            "b_dim_id": "model-serving-deployment-and-runtime-packaging",
            "b_name": "Model Serving Deployment and Packaging",
            "b_role": "__skill_focal__",
            "pair_kind": "intra_role",
            "reasoning": "A is about serving-system architecture: routing, autoscaling, canary rollout, and latency/throughput tradeoffs for online, batch, or streaming inference. Its exemplars include online inference, batch scoring, serving endpoints, and model routing. B is about deployment/runtime packaging: containers, managed inference runtimes, base images, GPU containers, entrypoints, dependencies, and platforms like TensorFlow Serving, TorchServe, Triton, BentoML, KServe, and Seldon Core. The shared word \"serving\" is misleading; A designs how inference traffic is handled, while B packages and deploys the model service into the runtime.",
            "similarity": 0.7076647206338962
          },
          {
            "a_dim_id": "model-serving-architecture",
            "a_name": "Model Serving Architecture",
            "a_role": "__skill_focal__",
            "b_dim_id": "model-serving-deployment-and-runtime-packaging",
            "b_name": "Model Serving Deployment and Runtime Packaging",
            "b_role": "MLOps Engineer",
            "pair_kind": "cross_role",
            "reasoning": "Dim A is about serving-system architecture: exposing inference, routing requests, autoscaling replicas, canary rollout, and latency/throughput tradeoffs. Its exemplars like online inference, batch scoring, serving endpoints, and model routing are runtime traffic concerns. Dim B is about deployment/runtime packaging: containerizing models and servers, choosing base images, GPU containers, runtime dependencies, and platforms like TensorFlow Serving, TorchServe, Triton, BentoML, KServe, and Seldon Core. Those exemplars are build/deploy concerns, not architecture design. The overlap on model serving is superficial.",
            "similarity": 0.7194474021288034
          },
          {
            "a_dim_id": "model-serving-deployment-and-runtime-packaging",
            "a_name": "Model Serving Deployment and Packaging",
            "a_role": "__skill_focal__",
            "b_dim_id": "model-serving-architecture",
            "b_name": "Model Serving Architecture",
            "b_role": "Machine Learning Engineer",
            "pair_kind": "cross_role",
            "reasoning": "Dim A is operational deployment: packaging model artifacts, containerized model servers, runtime dependencies, deployment manifests, and rollout automation. Its out_of_scope explicitly excludes request routing and service-level scaling. Dim B is architecture: hosting, routing, and scaling inference, including embedding model calls in services, sidecars, gateways, or dedicated serving layers. A\u2019s exemplar skills (model artifact packaging, containerized model server) differ from B\u2019s architectural concerns, so the overlap is only on the broad phrase \u201cmodel serving.\u201d",
            "similarity": 0.7282862514128082
          }
        ],
        "locked_dimensions": [
          {
            "description": "Patterns for hosting, routing, versioning, and scaling model inference in production systems, including online inference endpoints, batch scoring services, streaming inference, and serving topologies such as dedicated services, sidecars, gateways, or embedded serving layers. This skill covers request routing, autoscaling serving replicas, canary rollout for models, model version selection, and latency/throughput tradeoffs.",
            "exemplar_skills": [
              "Production Model Serving Architecture"
            ],
            "in_scope": "Skills, tools, and practices that belong under Production Model Serving Architecture for the target role, including items implied by the dimension rationale.",
            "name": "Production Model Serving Architecture",
            "out_of_scope": "Adjacent clusters explicitly not owned by Production Model Serving Architecture, including unrelated platforms, roles, and skill families per library policy.",
            "overlap_flags": [],
            "tentative_id": "d_merge_01"
          },
          {
            "description": "Operational deployment of trained models into online, batch, or streaming serving environments, including packaging model artifacts and model servers into containers or managed inference runtimes, managing runtime dependencies, defining deployment manifests and serving images, and coordinating rollout automation. Includes serving frameworks such as TensorFlow Serving, TorchServe, Triton, BentoML, KServe, and Seldon Core, plus Docker and GPU-enabled container concerns.",
            "exemplar_skills": [
              "Model Serving Deployment and Runtime Packaging"
            ],
            "in_scope": "Skills, tools, and practices that belong under Model Serving Deployment and Runtime Packaging for the target role, including items implied by the dimension rationale.",
            "name": "Model Serving Deployment and Runtime Packaging",
            "out_of_scope": "Adjacent clusters explicitly not owned by Model Serving Deployment and Runtime Packaging, including unrelated platforms, roles, and skill families per library policy.",
            "overlap_flags": [],
            "tentative_id": "d_merge_02"
          }
        ],
        "merge_log": [
          {
            "a_dim_id": "model-serving-architecture",
            "a_name": "Model Serving Architecture",
            "a_role": "__skill_focal__",
            "b_dim_id": "model-serving-architecture",
            "b_name": "Model Serving Architecture",
            "b_role": "Machine Learning Engineer",
            "into": "d_merge_01",
            "into_name": "Production Model Serving Architecture",
            "merged_from": [
              "model-serving-architecture",
              "model-serving-architecture"
            ],
            "pair_kind": "cross_role",
            "reasoning": "Both dimensions describe the same production inference-serving cluster: hosting, routing, versioning, and scaling model inference. Dim A covers online endpoints, batch scoring, streaming inference, request routing, autoscaling, canary rollout, and latency/throughput tradeoffs. Dim B says the same thing with serving topologies like services, sidecars, gateways, or dedicated layers. The exemplar skills in A all fit B, and B adds no distinct skill anchor beyond deployment form factors. This is a wording overlap, not a separate cluster.",
            "similarity": 0.912785730918857
          },
          {
            "a_dim_id": "model-serving-deployment-and-runtime-packaging",
            "a_name": "Model Serving Deployment and Packaging",
            "a_role": "__skill_focal__",
            "b_dim_id": "model-serving-deployment-and-runtime-packaging",
            "b_name": "Model Serving Deployment and Runtime Packaging",
            "b_role": "MLOps Engineer",
            "into": "d_merge_02",
            "into_name": "Model Serving Deployment and Runtime Packaging",
            "merged_from": [
              "model-serving-deployment-and-runtime-packaging",
              "model-serving-deployment-and-runtime-packaging"
            ],
            "pair_kind": "cross_role",
            "reasoning": "Both dims cover the same operational work: packaging trained models and deploying them into serving environments. A emphasizes model artifact packaging, containerized model servers, runtime dependency management, deployment manifests, and rollout automation for online/batch serving. B describes the same cluster with more platform examples (TensorFlow Serving, TorchServe, Triton, BentoML, KServe, Seldon Core) plus Docker/GPU image and runtime concerns. The exemplar skills overlap directly on model serving, packaging, containerized model server, deployment, dependencies, and rollout.",
            "similarity": 0.8622630010276275
          }
        ],
        "placed": {
          "name": "Model Serving",
          "placement_confidence": 0.92,
          "primary_dimension": "d_merge_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 2 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [
            "d_merge_02"
          ],
          "skill_id": "model-serving"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [],
          "related_to": [
            "mlops",
            "amazon-sagemaker",
            "mlflow",
            "large-language-models",
            "machine-learning",
            "ml",
            "ai-ml",
            "evaluation",
            "eval-design",
            "prompting"
          ],
          "requires": [],
          "skill_id": "model-serving",
          "suppress_on_match": []
        },
        "skill_id": "model-serving",
        "split_log": [],
        "typed": {
          "alternatives_considered": [
            "Concept: ruled out \u2014 it is more than a knowledge unit; it describes how systems are organized for inference.",
            "Methodology: ruled out \u2014 it is not a way of working or process.",
            "Service: ruled out \u2014 the term is generic and does not denote a specific managed capability within a platform."
          ],
          "confidence": 0.88,
          "name": "Model Serving",
          "reasoning": "Model Serving is fundamentally a system-shape pattern for exposing trained models for inference, so by the Architecture vs Concept rule it fits Architecture rather than a tool or service.",
          "skill_id": "model-serving",
          "subtype": "model_serving_architecture",
          "type": "Architecture"
        },
        "warnings": [
          "stage3_post_filter_dropped_catalog_only_locked_dims:40-\u003e2"
        ]
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Reinforcement Learning",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Reinforcement Learning",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Concept",
          "skill_nature": "CONCEPT",
          "sub_category": "machine_learning_paradigm",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "EMERGING"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "Reinforcement Learning is a well-established, specific ML paradigm with a distinctive full name. In typical JDs it is unlikely to be mistaken for another catalog skill."
          },
          "context_keywords": {
            "context_keywords": [
              "Q-learning",
              "policy gradient",
              "value function",
              "Markov decision process",
              "exploration-exploitation",
              "reward shaping",
              "temporal difference",
              "actor-critic",
              "deep reinforcement learning",
              "epsilon-greedy",
              "Monte Carlo methods",
              "experience replay",
              "gymnasium",
              "multi-armed bandit",
              "Bellman equation"
            ]
          },
          "maturity": {
            "confidence": 0.84,
            "maturity": "emerging",
            "reasoning": "Appears increasingly in ML/AI job descriptions and vendor roadmaps, especially for robotics and LLM alignment, but remains far less common than supervised learning roles."
          },
          "skill_id": "reinforcement-learning",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Methods for learning policies through trial-and-error interaction with an environment to maximize cumulative reward. This includes value-based, policy-based, and actor-critic approaches used in machine learning systems.",
            "exemplar_skills": [
              "Reinforcement Learning",
              "Q-learning",
              "SARSA",
              "Policy Gradients",
              "Actor-Critic",
              "Temporal-Difference Learning",
              "Markov Decision Processes"
            ],
            "in_scope": "Reinforcement Learning, Q-learning, SARSA, policy gradients, actor-critic methods, exploration-exploitation, reward shaping, Markov decision processes, temporal-difference learning, multi-armed bandits, offline reinforcement learning, model-free RL, model-based RL",
            "name": "Reinforcement Learning",
            "out_of_scope": "Supervised learning for labeled prediction tasks, unsupervised clustering or dimensionality reduction, model serving and deployment, experiment design for A/B tests, which belong to other dimensions",
            "overlap_flags": [
              {
                "reason": "RL often uses experiments to evaluate policies, but the core skill is sequential decision-making rather than causal test design.",
                "with_dim_id": "experiment-design-and-analysis",
                "with_dim_name": null,
                "with_role": "Data Scientist"
              },
              {
                "reason": "Trained RL policies may be deployed like other models, but deployment is operational rather than the learning method itself.",
                "with_dim_id": "model-serving-deployment-and-runtime-packaging",
                "with_dim_name": null,
                "with_role": "MLOps Engineer, Machine Learning Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          }
        ],
        "merge_log": [],
        "placed": {
          "name": "Reinforcement Learning",
          "placement_confidence": 0.92,
          "primary_dimension": "d_init_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 1 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [],
          "skill_id": "reinforcement-learning"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [
            "machine-learning"
          ],
          "related_to": [
            "artificial-intelligence",
            "ai-ml",
            "ai",
            "ml",
            "algorithms",
            "evaluation",
            "mlops",
            "retrieval",
            "hybrid-retrieval",
            "reranking"
          ],
          "requires": [],
          "skill_id": "reinforcement-learning",
          "suppress_on_match": []
        },
        "skill_id": "reinforcement-learning",
        "split_log": [],
        "typed": {
          "alternatives_considered": [],
          "confidence": 0.97,
          "name": "Reinforcement Learning",
          "reasoning": "Reinforcement Learning is fundamentally a named knowledge unit about how agents learn via rewards, so by the Concept vs Methodology rule it is a Concept rather than a tool or process.",
          "skill_id": "reinforcement-learning",
          "subtype": "machine_learning_paradigm",
          "type": "Concept"
        },
        "warnings": []
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Graph Neural Networks",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Project Delivery and Coordination",
            "id": 366,
            "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
            "slug": "d_init_02",
            "source": "db"
          },
          "input_skill": "Graph Neural Networks",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Graph Neural Networks",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Concept",
          "skill_nature": "CONCEPT",
          "sub_category": "machine_learning_model",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "EMERGING"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "Graph Neural Networks is a specific ML model family with a clear, standard name. In typical JDs it is unlikely to be mistaken for a different catalog skill."
          },
          "context_keywords": {
            "context_keywords": [
              "message passing",
              "node embeddings",
              "edge features",
              "graph convolution",
              "GNN",
              "GCN",
              "GraphSAGE",
              "GAT",
              "heterogeneous graphs",
              "link prediction",
              "node classification",
              "graph pooling",
              "attention mechanism",
              "PyTorch Geometric",
              "DGL"
            ]
          },
          "maturity": {
            "confidence": 0.84,
            "maturity": "emerging",
            "reasoning": "Graph Neural Networks appear in growing numbers of ML/AI job postings and research-to-production tooling, but they are still far less common than core deep learning stacks like PyTorch/TensorFlow."
          },
          "skill_id": "graph-neural-networks",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Techniques for building and training neural models that operate on graph-structured data. This includes message passing, node/edge embeddings, and graph-level prediction, which are the core ideas behind Graph Neural Networks.",
            "exemplar_skills": [
              "Graph Neural Networks",
              "Graph Convolutional Networks",
              "Graph Attention Networks",
              "Message Passing Neural Networks",
              "Node Classification",
              "Link Prediction"
            ],
            "in_scope": "Graph Neural Networks, message passing neural networks, graph convolutional networks, graph attention networks, node classification, link prediction, graph classification, node embeddings, edge embeddings, graph sampling, heterogeneous graphs",
            "name": "Graph Neural Network Modeling",
            "out_of_scope": "General deep learning architectures for images or text, graph databases and storage systems, distributed graph processing infrastructure, model serving and deployment concerns",
            "overlap_flags": [
              {
                "reason": "GNN work is often implemented in Python or similar analytical languages, but the dimension here is the graph-modeling technique rather than the coding language.",
                "with_dim_id": "analytical-programming-languages",
                "with_dim_name": null,
                "with_role": "Data Analyst, Data Scientist"
              }
            ],
            "tentative_id": "d_init_01"
          },
          {
            "description": "Learning useful vector representations from graph structure and attributes for downstream machine learning tasks. This dimension covers embedding-based methods and representation learning approaches that often include, but are broader than, specific GNN architectures.",
            "exemplar_skills": [
              "Graph Neural Networks",
              "Graph Embeddings",
              "Node Representation Learning",
              "Edge Representation Learning",
              "Self-Supervised Graph Learning",
              "Contrastive Graph Learning"
            ],
            "in_scope": "Graph Neural Networks, graph embeddings, node representation learning, edge representation learning, self-supervised graph learning, contrastive graph learning, relational learning, subgraph encoding",
            "name": "Graph Representation Learning",
            "out_of_scope": "Graph databases, graph algorithms for shortest path or traversal, general representation learning for non-graph data, serving and deployment of trained models",
            "overlap_flags": [
              {
                "reason": "Both can involve learned embeddings, but this dimension is specifically about graph-structured representation learning rather than retrieval-augmented context selection.",
                "with_dim_id": "context-management-and-retrieval",
                "with_dim_name": null,
                "with_role": "AI Engineer"
              }
            ],
            "tentative_id": "d_init_02"
          }
        ],
        "merge_log": [],
        "placed": {
          "name": "Graph Neural Networks",
          "placement_confidence": 0.92,
          "primary_dimension": "d_init_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 2 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [
            "d_init_02"
          ],
          "skill_id": "graph-neural-networks"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [
            "machine-learning",
            "ai-ml"
          ],
          "related_to": [
            "large-language-models",
            "vision-language-models",
            "subgraphs",
            "the-graph",
            "quicknode",
            "cloud-native",
            "nosql",
            "multimodal-document-understanding",
            "pyspark",
            "data-pipelines"
          ],
          "requires": [],
          "skill_id": "graph-neural-networks",
          "suppress_on_match": []
        },
        "skill_id": "graph-neural-networks",
        "split_log": [],
        "typed": {
          "alternatives_considered": [],
          "confidence": 0.96,
          "name": "Graph Neural Networks",
          "reasoning": "Graph Neural Networks are a named machine-learning model family/knowledge unit rather than software you run or build inside, so they fit the Concept type.",
          "skill_id": "graph-neural-networks",
          "subtype": "machine_learning_model",
          "type": "Concept"
        },
        "warnings": []
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "RAG",
          "alias_type": "CANONICAL",
          "id": 3616,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 2,
        "display_name": "RAG",
        "id": 2659,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "CONCEPT",
        "slug": "rag",
        "sub_category_id": 2172,
        "typical_lifespan": "EVERGREEN",
        "volatility": "EMERGING"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Cloud Service Integration Patterns",
            "id": 188,
            "rationale": "Covers how cloud services and workloads connect through APIs, events, shared services, and integration boundaries. This cluster is coherent because architects must define interaction patterns that preserve decoupling, security, and operability.",
            "slug": "cloud-service-integration-patterns",
            "source": "db"
          },
          "input_skill": "RAG",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "Cloud Architect",
              "id": 11,
              "rationale": null,
              "role_archetype": null,
              "slug": "cloud-architect",
              "source": "db"
            }
          ]
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Context Management and Retrieval",
            "id": 264,
            "rationale": "Preparing, selecting, and packaging context for model calls so responses stay relevant and grounded. This is a distinct cluster because AI features often depend on what information is included, summarized, or retrieved at call time.",
            "slug": "context-management-and-retrieval",
            "source": "db"
          },
          "input_skill": "RAG",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "AI Engineer",
              "id": 12,
              "rationale": null,
              "role_archetype": null,
              "slug": "ai-engineer",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "RAG",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": null,
            "display_name": "Experimental Design, Testing, and Causal Analysis",
            "id": null,
            "rationale": "Designing controlled experiments and analyzing results to estimate causal impact. This includes A/B tests, randomized experiments, hypothesis testing, metric selection, statistical significance, treatment/control design, power analysis, causal inference, and evaluating model or product variants through rigorous comparison. It covers the experimental method and interpretation of outcomes, not the mechanics of model training or data preparation.",
            "slug": "d_merge_01",
            "source": "llm"
          },
          "input_skill": "Fine-tuning",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Fine-tuning",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Fine-tuning",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Methodology",
          "skill_nature": "METHODOLOGY",
          "sub_category": "model_fine_tuning_methodology",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "STABLE"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "\u201cFine-tuning\u201d is a standard ML methodology term and is usually clear in JDs. It\u2019s unlikely to be reasonably mistaken for a different catalog skill."
          },
          "context_keywords": {
            "context_keywords": [
              "transfer learning",
              "pretrained model",
              "checkpoint",
              "hyperparameter tuning",
              "learning rate scheduler",
              "gradient accumulation",
              "mixed precision",
              "LoRA",
              "PEFT",
              "adapter layers",
              "instruction tuning",
              "domain adaptation",
              "catastrophic forgetting",
              "SFT",
              "parameter-efficient fine-tuning"
            ]
          },
          "maturity": {
            "confidence": 0.93,
            "maturity": "well_known",
            "reasoning": "Widely listed in ML/LLM job descriptions and vendor docs (OpenAI, Hugging Face) as a standard customization approach; strong GitHub/tutorial volume and active tooling ecosystem signal broad adoption."
          },
          "skill_id": "fine-tuning",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Designing controlled experiments and analyzing results to estimate causal impact. This includes A/B tests, randomized experiments, hypothesis testing, metric selection, statistical significance, treatment/control design, power analysis, causal inference, and evaluating model or product variants through rigorous comparison. It covers the experimental method and interpretation of outcomes, not the mechanics of model training or data preparation.",
            "exemplar_skills": [
              "Experimental Design, Testing, and Causal Analysis"
            ],
            "in_scope": "Skills, tools, and practices that belong under Experimental Design, Testing, and Causal Analysis for the target role, including items implied by the dimension rationale.",
            "name": "Experimental Design, Testing, and Causal Analysis",
            "out_of_scope": "Adjacent clusters explicitly not owned by Experimental Design, Testing, and Causal Analysis, including unrelated platforms, roles, and skill families per library policy.",
            "overlap_flags": [],
            "tentative_id": "d_merge_01"
          },
          {
            "description": "Adapting a pretrained model to a target task or domain using supervised, instruction, or preference-based updates. This skill belongs here because it is about modifying model weights and training behavior, not deploying or serving the model.",
            "exemplar_skills": [
              "Fine-tuning",
              "Supervised fine-tuning",
              "Instruction tuning",
              "LoRA",
              "Adapter tuning",
              "Transfer learning"
            ],
            "in_scope": "Fine-tuning, supervised fine-tuning, instruction tuning, domain adaptation, parameter-efficient fine-tuning, LoRA, adapters, prompt-tuning, transfer learning, continued pretraining, model adaptation",
            "name": "Model Fine-Tuning",
            "out_of_scope": "Model serving, inference routing, container packaging, batch scoring pipelines, experiment analysis, data labeling operations, which belong to serving, pipelines, or experimentation dimensions",
            "overlap_flags": [
              {
                "reason": "Fine-tuning work is often evaluated with experiments, but this dimension covers the training adaptation itself rather than causal analysis or A/B testing.",
                "with_dim_id": "experiment-design-and-analysis",
                "with_dim_name": null,
                "with_role": "Data Scientist"
              },
              {
                "reason": "Fine-tuning may consume curated datasets, but data movement and feature refresh belong to pipeline operations, not model adaptation.",
                "with_dim_id": "inference-data-pipelines",
                "with_dim_name": null,
                "with_role": "MLOps Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          }
        ],
        "merge_log": [
          {
            "a_dim_id": "experiment-design-and-analysis",
            "a_name": "Experiment Design and Analysis",
            "a_role": "__skill_focal__",
            "b_dim_id": "experiment-design-and-analysis",
            "b_name": "Experiment Design and Analysis",
            "b_role": "Data Scientist",
            "into": "d_merge_01",
            "into_name": "Experimental Design, Testing, and Causal Analysis",
            "merged_from": [
              "experiment-design-and-analysis",
              "experiment-design-and-analysis"
            ],
            "pair_kind": "cross_role",
            "reasoning": "These are the same conceptual cluster with identical names and essentially identical descriptions: both focus on designing controlled experiments and analyzing outcomes to estimate causal impact. Dim A\u2019s exemplars (A/B testing, hypothesis testing, power analysis, causal inference, metric selection, and even fine-tuning evaluation) all fit the same experimental-method skill cluster described in Dim B. The extra note in Dim A about fine-tuning overlap is just a boundary clarification, not a different skill area. Since Dim B has no distinct scope content and the pair is cross_role but still semantically identical, this is not a role-specific distinction; it is the same experiment-design skill set.",
            "similarity": 0.809223867249764
          }
        ],
        "placed": {
          "name": "Fine-tuning",
          "placement_confidence": 0.92,
          "primary_dimension": "d_merge_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 2 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [
            "d_init_01"
          ],
          "skill_id": "fine-tuning"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [],
          "related_to": [
            "prompt-engineering",
            "prompting",
            "evaluation",
            "eval-design",
            "evaluation-design",
            "failure-analysis",
            "mlops",
            "reranking",
            "agent-tooling",
            "document-processing"
          ],
          "requires": [],
          "skill_id": "fine-tuning",
          "suppress_on_match": []
        },
        "skill_id": "fine-tuning",
        "split_log": [],
        "typed": {
          "alternatives_considered": [],
          "confidence": 0.9,
          "name": "Fine-tuning",
          "reasoning": "By the Concept vs Methodology rule, fine-tuning is a way of working to adapt a model rather than a standalone knowledge unit or system component.",
          "skill_id": "fine-tuning",
          "subtype": "model_fine_tuning_methodology",
          "type": "Methodology"
        },
        "warnings": [
          "stage3_post_filter_dropped_catalog_only_locked_dims:41-\u003e2"
        ]
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [
        {
          "alias_text": "Prompt Engineering",
          "alias_type": "CANONICAL",
          "id": 3658,
          "is_primary": false,
          "match_strategy": "CASE_INSENSITIVE"
        }
      ],
      "canonical": {
        "category_id": 7,
        "display_name": "Prompt Engineering",
        "id": 2683,
        "is_also_category": false,
        "is_extractable": true,
        "skill_nature": "METHODOLOGY",
        "slug": "prompt-engineering",
        "sub_category_id": 2191,
        "typical_lifespan": "EVERGREEN",
        "volatility": "EMERGING"
      },
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Context Management and Retrieval",
            "id": 264,
            "rationale": "Preparing, selecting, and packaging context for model calls so responses stay relevant and grounded. This is a distinct cluster because AI features often depend on what information is included, summarized, or retrieved at call time.",
            "slug": "context-management-and-retrieval",
            "source": "db"
          },
          "input_skill": "Prompt Engineering",
          "llm_role": null,
          "roles_from_db": [
            {
              "display_name": "AI Engineer",
              "id": 12,
              "rationale": null,
              "role_archetype": null,
              "slug": "ai-engineer",
              "source": "db"
            }
          ]
        }
      ],
      "input_skill": "Prompt Engineering",
      "matched_via": "alias",
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": null,
      "source_tag": "db",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Classification",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Classification",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Concept",
          "skill_nature": "CONCEPT",
          "sub_category": "classification_concept",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "STABLE"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": true,
            "confused_with": [
              "machine_learning",
              "supervised_learning"
            ],
            "reasoning": "\"Classification\" is a broad ML concept and in JDs can be used loosely to mean machine learning or supervised learning tasks. A parser could confuse it with those related catalog skills."
          },
          "context_keywords": {
            "context_keywords": [
              "supervised learning",
              "labeling",
              "training set",
              "test set",
              "confusion matrix",
              "precision",
              "recall",
              "F1 score",
              "ROC curve",
              "AUC",
              "logistic regression",
              "decision tree",
              "random forest",
              "SVM",
              "multiclass"
            ]
          },
          "maturity": {
            "confidence": 0.93,
            "maturity": "well_known",
            "reasoning": "Broadly used across ML/data roles; appears routinely in job descriptions for supervised learning, fraud/spam detection, and document tagging, with strong GitHub/tutorial ecosystem signal."
          },
          "skill_id": "classification",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Supervised learning tasks that assign inputs to discrete labels or classes. This fits the target skill because classification is a core ML problem type used to predict categories, ranks, or binary outcomes from features.",
            "exemplar_skills": [
              "Classification",
              "binary classification",
              "multiclass classification",
              "logistic regression",
              "decision tree classification",
              "random forest classification",
              "support vector machine classification"
            ],
            "in_scope": "Classification, binary classification, multiclass classification, multilabel classification, logistic regression, decision trees, random forests, support vector machines, naive Bayes, class imbalance handling, threshold tuning, precision/recall tradeoffs",
            "name": "Machine Learning Classification",
            "out_of_scope": "Regression and forecasting, clustering and dimensionality reduction, model serving and deployment, feature engineering pipelines, which belong to other model-development or operational dimensions",
            "overlap_flags": [
              {
                "reason": "Classification model evaluation often uses controlled experiments and metric comparison, but the core skill is model formulation rather than experiment design.",
                "with_dim_id": "experiment-design-and-analysis",
                "with_dim_name": null,
                "with_role": "Data Scientist"
              },
              {
                "reason": "Classification models are often deployed for inference, yet serving concerns are separate from learning the classification task itself.",
                "with_dim_id": "model-serving-architecture",
                "with_dim_name": null,
                "with_role": "Machine Learning Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          }
        ],
        "merge_log": [],
        "placed": {
          "name": "Classification",
          "placement_confidence": 0.92,
          "primary_dimension": "d_init_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 1 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [],
          "skill_id": "classification"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [],
          "related_to": [
            "evaluation",
            "ocr",
            "ml",
            "machine-learning",
            "reranking",
            "eval-design",
            "data-modeling",
            "sql",
            "data-structures",
            "code-review"
          ],
          "requires": [],
          "skill_id": "classification",
          "suppress_on_match": []
        },
        "skill_id": "classification",
        "split_log": [],
        "typed": {
          "alternatives_considered": [],
          "confidence": 0.95,
          "name": "Classification",
          "reasoning": "Classification is a named knowledge unit about assigning items to categories, so by the Concept vs Methodology rule it is a Concept rather than a process or tool.",
          "skill_id": "classification",
          "subtype": "classification_concept",
          "type": "Concept"
        },
        "warnings": []
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Ranking",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Ranking",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Concept",
          "skill_nature": "CONCEPT",
          "sub_category": "ranking_concept",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "STABLE"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "\u201cRanking\u201d is a broad concept, but in JDs it usually appears in clear context (e.g., search ranking, model ranking, leaderboard ranking). It is not a short acronym or vendor/product name likely to be mistaken for a distinct catalog skill."
          },
          "context_keywords": {
            "context_keywords": [
              "relevance scoring",
              "learning to rank",
              "pairwise ranking",
              "listwise ranking",
              "pointwise ranking",
              "NDCG",
              "MAP",
              "MRR",
              "A/B testing",
              "click-through rate",
              "search results",
              "recommendation systems",
              "LTR",
              "feature engineering",
              "candidate generation"
            ]
          },
          "maturity": {
            "confidence": 0.84,
            "maturity": "well_known",
            "reasoning": "Ranking is a core concept in search/recommendation and ML roles; it appears broadly in job descriptions for search relevance, recommender systems, and IR, with strong market demand across major platforms."
          },
          "skill_id": "ranking",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Methods for ordering items by predicted importance, relevance, or utility. This fits machine learning work when the skill is about producing ranked outputs for search, recommendations, retrieval, or prioritization.",
            "exemplar_skills": [
              "Ranking",
              "learning to rank",
              "relevance scoring",
              "pairwise ranking",
              "listwise ranking",
              "NDCG",
              "MAP",
              "reranking"
            ],
            "in_scope": "Ranking, learning-to-rank, relevance scoring, pairwise ranking, listwise ranking, score calibration for ordering, top-k selection, reranking, ranking metrics such as NDCG and MAP",
            "name": "Ranking and Relevance Scoring",
            "out_of_scope": "Classification labels, regression targets, clustering, experiment design and A/B testing, model serving deployment, feature engineering for ranking inputs",
            "overlap_flags": [
              {
                "reason": "Ranking systems are often evaluated with offline and online experiments, but that dimension owns causal test design and analysis rather than the ranking objective itself.",
                "with_dim_id": "experiment-design-and-analysis",
                "with_dim_name": null,
                "with_role": "Data Scientist"
              },
              {
                "reason": "Production ranking models may be served online, but this dimension focuses on the ranking method and metrics, not deployment patterns.",
                "with_dim_id": "model-serving-architecture",
                "with_dim_name": null,
                "with_role": "Machine Learning Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          }
        ],
        "merge_log": [],
        "placed": {
          "name": "Ranking",
          "placement_confidence": 0.92,
          "primary_dimension": "d_init_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 1 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [],
          "skill_id": "ranking"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [],
          "related_to": [
            "reranking",
            "evaluation",
            "queueing",
            "retrieval",
            "hybrid-retrieval",
            "sql",
            "ocr",
            "code-review",
            "tally",
            "efficiency"
          ],
          "requires": [],
          "skill_id": "ranking",
          "suppress_on_match": []
        },
        "skill_id": "ranking",
        "split_log": [],
        "typed": {
          "alternatives_considered": [
            "Methodology: ruled out \u2014 ranking is not a way of working or process.",
            "Architecture: ruled out \u2014 ranking is not a system-shape pattern."
          ],
          "confidence": 0.88,
          "name": "Ranking",
          "reasoning": "Ranking is a named knowledge unit about ordering or prioritization, so by the Concept vs Methodology rule it is a Concept rather than a process or system shape.",
          "skill_id": "ranking",
          "subtype": "ranking_concept",
          "type": "Concept"
        },
        "warnings": []
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Fraud Detection",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Project Delivery and Coordination",
            "id": 366,
            "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
            "slug": "d_init_02",
            "source": "db"
          },
          "input_skill": "Fraud Detection",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Fraud Detection",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Domain",
          "skill_nature": "CONCEPT",
          "sub_category": "fraud_detection",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "STABLE"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "Fraud Detection is a clear domain term in JDs and is unlikely to be mistaken for a different catalog skill; it is more often a task area than an ambiguous acronym or product name."
          },
          "context_keywords": {
            "context_keywords": [
              "anomaly detection",
              "transaction monitoring",
              "chargeback",
              "AML",
              "KYC",
              "identity verification",
              "risk scoring",
              "rules engine",
              "case management",
              "false positives",
              "velocity checks",
              "device fingerprinting",
              "graph analytics",
              "suspicious activity",
              "sanctions screening"
            ]
          },
          "maturity": {
            "confidence": 0.92,
            "maturity": "well_known",
            "reasoning": "Fraud detection is broadly listed in fintech, payments, and banking job descriptions, and major vendors like Stripe, Adyen, and AWS offer dedicated fraud products, indicating strong market demand."
          },
          "skill_id": "fraud-detection",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Methods for identifying suspicious or deceptive activity using statistical, machine learning, and rule-based signals. Fraud Detection belongs here because it is a core applied ML problem focused on scoring, ranking, and flagging anomalous transactions or behaviors.",
            "exemplar_skills": [
              "Fraud Detection",
              "fraud scoring",
              "chargeback prediction",
              "anomaly detection",
              "transaction risk scoring",
              "account takeover detection"
            ],
            "in_scope": "Fraud Detection, fraud scoring, anomaly-based fraud signals, supervised fraud classification, transaction risk scoring, account takeover detection, chargeback prediction, rule-based fraud heuristics, feature engineering for fraud models, fraud alert prioritization",
            "name": "Fraud Detection Modeling",
            "out_of_scope": "Identity login hardening and MFA belong to authentication-and-session-security, incident handling after a fraud alert belongs to incident-response-and-containment, payment gateway integration belongs to cloud-service-integration-patterns",
            "overlap_flags": [
              {
                "reason": "Fraud signals often use login and session telemetry, but the identity controls themselves are owned by authentication.",
                "with_dim_id": "authentication-and-session-security",
                "with_dim_name": null,
                "with_role": "Cybersecurity Engineer"
              },
              {
                "reason": "Fraud detection can trigger investigations, but containment and response workflows are a separate operational skill.",
                "with_dim_id": "incident-response-and-containment",
                "with_dim_name": null,
                "with_role": "Cybersecurity Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          },
          {
            "description": "Techniques for detecting unusual patterns and assigning risk scores to events, users, or transactions. Fraud Detection fits this dimension when the emphasis is on spotting deviations, ranking suspicious cases, and calibrating thresholds rather than domain-specific fraud policy.",
            "exemplar_skills": [
              "Fraud Detection",
              "anomaly detection",
              "risk scoring",
              "outlier detection",
              "alert ranking",
              "threshold tuning"
            ],
            "in_scope": "Fraud Detection, anomaly detection, outlier scoring, risk scoring, threshold tuning, alert ranking, behavioral deviation analysis, unsupervised detection, semi-supervised detection",
            "name": "Anomaly Detection and Risk Scoring",
            "out_of_scope": "Case management, investigator workflows, and remediation belong to incident-response-and-containment, payment or banking domain rules belong to domain-specific business logic, model deployment concerns belong to model-serving-deployment-and-runtime-packaging",
            "overlap_flags": [
              {
                "reason": "Fraud models often need offline evaluation and threshold calibration, but causal experimentation is a separate dimension.",
                "with_dim_id": "experiment-design-and-analysis",
                "with_dim_name": null,
                "with_role": "Data Scientist"
              },
              {
                "reason": "Fraud scoring is frequently deployed online, but serving patterns are distinct from the detection logic itself.",
                "with_dim_id": "model-serving-architecture",
                "with_dim_name": null,
                "with_role": "Machine Learning Engineer"
              }
            ],
            "tentative_id": "d_init_02"
          }
        ],
        "merge_log": [],
        "placed": {
          "name": "Fraud Detection",
          "placement_confidence": 0.92,
          "primary_dimension": "d_init_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 2 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [
            "d_init_02"
          ],
          "skill_id": "fraud-detection"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [],
          "related_to": [
            "failure-analysis",
            "ai-ml",
            "code-review",
            "hybrid-retrieval",
            "document-processing",
            "data-modeling",
            "data-ingestion",
            "data-pipelines",
            "evaluation",
            "workflow-automation"
          ],
          "requires": [
            "machine-learning",
            "security",
            "authentication",
            "nonce-handling",
            "merkle-proofs"
          ],
          "skill_id": "fraud-detection",
          "suppress_on_match": []
        },
        "skill_id": "fraud-detection",
        "split_log": [],
        "typed": {
          "alternatives_considered": [],
          "confidence": 0.98,
          "name": "Fraud Detection",
          "reasoning": "Fraud Detection is a vertical problem-space body of knowledge, so by the Domain rule it is not a tool, method, or architecture.",
          "skill_id": "fraud-detection",
          "subtype": "fraud_detection",
          "type": "Domain"
        },
        "warnings": []
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    },
    {
      "aliases_in_db": [],
      "canonical": null,
      "dimensions": [
        {
          "dimension": {
            "difficulty_hint": null,
            "display_name": "Experiment Design, Causal Inference, and Analysis",
            "id": null,
            "rationale": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This includes A/B testing, randomized controlled trials, hypothesis testing, confidence intervals, causal inference, uplift analysis, backtesting, and treatment effect estimation. The focus is on rigorous causal measurement of interventions, treatments, or product changes, not on risk modeling, deployment, or reporting.",
            "slug": "d_merge_01",
            "source": "llm"
          },
          "input_skill": "Risk Modeling",
          "llm_role": null,
          "roles_from_db": []
        },
        {
          "dimension": {
            "difficulty_hint": "well_known",
            "display_name": "Version Control Systems",
            "id": 365,
            "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
            "slug": "d_init_01",
            "source": "db"
          },
          "input_skill": "Risk Modeling",
          "llm_role": null,
          "roles_from_db": []
        }
      ],
      "input_skill": "Risk Modeling",
      "matched_via": null,
      "new_alias_persisted": false,
      "new_alias_text": null,
      "new_skill_meta": {
        "derived": {
          "category": "Concept",
          "skill_nature": "CONCEPT",
          "sub_category": "risk_modeling",
          "typical_lifespan": "EVERGREEN",
          "version_strategy": "NOT_APPLICABLE",
          "volatility": "STABLE"
        },
        "enrichment": {
          "ambiguity": {
            "ambiguity_flag": false,
            "confused_with": [],
            "reasoning": "\u201cRisk Modeling\u201d is a fairly specific finance/risk concept and is unlikely to be mistaken for a different catalog skill in a typical job description."
          },
          "context_keywords": {
            "context_keywords": [
              "Monte Carlo simulation",
              "stress testing",
              "scenario analysis",
              "VaR",
              "expected shortfall",
              "sensitivity analysis",
              "loss distribution",
              "credit risk",
              "market risk",
              "operational risk",
              "probability of default",
              "exposure at default",
              "loss given default",
              "capital adequacy",
              "backtesting"
            ]
          },
          "maturity": {
            "confidence": 0.86,
            "maturity": "well_known",
            "reasoning": "Common in finance, insurance, and fintech JDs; roles often require risk modeling, credit/market risk, and stress testing, with strong hiring demand across banks and regulators."
          },
          "skill_id": "risk-modeling",
          "vendor_license": {
            "confidence": 0.99,
            "license": null,
            "vendor": null,
            "year_introduced": null
          },
          "versioning": {
            "current_version": null,
            "version_aliases": {},
            "versioned": false
          }
        },
        "keep_log": [],
        "locked_dimensions": [
          {
            "description": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This includes A/B testing, randomized controlled trials, hypothesis testing, confidence intervals, causal inference, uplift analysis, backtesting, and treatment effect estimation. The focus is on rigorous causal measurement of interventions, treatments, or product changes, not on risk modeling, deployment, or reporting.",
            "exemplar_skills": [
              "Experiment Design, Causal Inference, and Analysis"
            ],
            "in_scope": "Skills, tools, and practices that belong under Experiment Design, Causal Inference, and Analysis for the target role, including items implied by the dimension rationale.",
            "name": "Experiment Design, Causal Inference, and Analysis",
            "out_of_scope": "Adjacent clusters explicitly not owned by Experiment Design, Causal Inference, and Analysis, including unrelated platforms, roles, and skill families per library policy.",
            "overlap_flags": [],
            "tentative_id": "d_merge_01"
          },
          {
            "description": "Building quantitative or statistical models to estimate uncertainty, loss, exposure, or likelihood of adverse outcomes. This fits the target skill because it focuses on representing risk factors, assumptions, and scenario outcomes in a structured model.",
            "exemplar_skills": [
              "Risk Modeling",
              "Credit Risk Modeling",
              "Operational Risk Modeling",
              "Fraud Risk Scoring",
              "Default Probability Estimation",
              "Stress Testing",
              "Monte Carlo Simulation"
            ],
            "in_scope": "Risk Modeling, credit risk models, market risk models, operational risk models, fraud risk scoring, default probability estimation, loss given default, exposure at default, scenario analysis, Monte Carlo simulation, stress testing, risk factor modeling",
            "name": "Risk Modeling",
            "out_of_scope": "Experiment Design and Analysis for causal treatment comparison, Model Serving Deployment and Runtime Packaging for deploying trained models, reporting and dashboard development for presenting results rather than building the model",
            "overlap_flags": [
              {
                "reason": "Risk models may be validated with experiments or backtesting, but the core work is estimating uncertainty and loss rather than causal inference.",
                "with_dim_id": "experiment-design-and-analysis",
                "with_dim_name": null,
                "with_role": "Data Scientist"
              },
              {
                "reason": "Risk models can be deployed like other ML models, but serving is operational packaging rather than the modeling discipline itself.",
                "with_dim_id": "model-serving-deployment-and-runtime-packaging",
                "with_dim_name": null,
                "with_role": "MLOps Engineer, Machine Learning Engineer"
              }
            ],
            "tentative_id": "d_init_01"
          }
        ],
        "merge_log": [
          {
            "a_dim_id": "experiment-design-and-analysis",
            "a_name": "Experiment Design and Analysis",
            "a_role": "__skill_focal__",
            "b_dim_id": "experiment-design-and-analysis",
            "b_name": "Experiment Design and Analysis",
            "b_role": "Data Scientist",
            "into": "d_merge_01",
            "into_name": "Experiment Design, Causal Inference, and Analysis",
            "merged_from": [
              "experiment-design-and-analysis",
              "experiment-design-and-analysis"
            ],
            "pair_kind": "cross_role",
            "reasoning": "Both dimensions describe the same conceptual cluster: designing controlled experiments and analyzing results to estimate causal impact. Dim A explicitly includes A/B testing, randomized controlled trials, causal inference, hypothesis testing, confidence intervals, and treatment effect estimation, and its description says the focus is causal measurement rather than risk estimation. Dim B has the same name and essentially the same description, emphasizing comparing interventions, treatments, or product changes with rigor. There is no substantive difference in skill content; the cross-role difference is only in framing, not in the underlying cluster. The exemplar skills in A map directly to B\u0027s description, so these should be unified rather than split or kept separate.",
            "similarity": 0.8247931716503335
          }
        ],
        "placed": {
          "name": "Risk Modeling",
          "placement_confidence": 0.92,
          "primary_dimension": "d_merge_01",
          "reasoning": "Deterministic JD placement: locked_dimensions has 2 dimension(s) from skill-driven dimension generation after reconciliation; primary_dimension is the first locked dim.",
          "secondary_dimensions": [
            "d_init_01"
          ],
          "skill_id": "risk-modeling"
        },
        "relationships": {
          "child_skills": [],
          "parent_skills": [],
          "related_to": [
            "data-modeling",
            "capacity-forecasting",
            "failure-analysis",
            "machine-learning",
            "evaluation-design",
            "ml",
            "large-language-models",
            "eval-design",
            "rollback-planning",
            "ai-ml"
          ],
          "requires": [],
          "skill_id": "risk-modeling",
          "suppress_on_match": []
        },
        "skill_id": "risk-modeling",
        "split_log": [],
        "typed": {
          "alternatives_considered": [],
          "confidence": 0.93,
          "name": "Risk Modeling",
          "reasoning": "Risk Modeling is fundamentally a named knowledge unit about assessing and quantifying risk, so under the Concept vs Methodology rule it fits Concept rather than a way of working.",
          "skill_id": "risk-modeling",
          "subtype": "risk_modeling",
          "type": "Concept"
        },
        "warnings": [
          "stage3_post_filter_dropped_catalog_only_locked_dims:41-\u003e2"
        ]
      },
      "source_tag": "llm",
      "was_in_llm_skills": true
    }
  ],
  "unmatched_skills": [
    "Feature Stores",
    "Model Serving",
    "Reinforcement Learning",
    "Graph Neural Networks",
    "Fine-tuning",
    "Classification",
    "Ranking",
    "Fraud Detection",
    "Risk Modeling"
  ]
}

{
  "chosen_role": {
    "display_name": "Machine Learning Engineer",
    "id": 10,
    "rationale": "The primary skills include Machine Learning, Python, and various deep learning frameworks, aligning well with the role of a Machine Learning Engineer.",
    "role_archetype": null,
    "slug": "machine-learning-engineer",
    "source": "db"
  },
  "chosen_role_resolution": "in_db",
  "final_input_skills": [
    {
      "skill": "Machine Learning",
      "tag": "in_db"
    },
    {
      "skill": "Python",
      "tag": "in_db"
    },
    {
      "skill": "PyTorch",
      "tag": "in_db"
    },
    {
      "skill": "TensorFlow",
      "tag": "in_db"
    },
    {
      "skill": "pandas",
      "tag": "in_db"
    },
    {
      "skill": "NumPy",
      "tag": "in_db"
    },
    {
      "skill": "OCR",
      "tag": "in_db"
    },
    {
      "skill": "A/B Testing",
      "tag": "in_db"
    },
    {
      "skill": "Feature Stores",
      "tag": "new"
    },
    {
      "skill": "Model Serving",
      "tag": "new"
    },
    {
      "skill": "Reinforcement Learning",
      "tag": "new"
    },
    {
      "skill": "Graph Neural Networks",
      "tag": "new"
    },
    {
      "skill": "RAG",
      "tag": "in_db"
    },
    {
      "skill": "Fine-tuning",
      "tag": "new"
    },
    {
      "skill": "Prompt Engineering",
      "tag": "in_db"
    },
    {
      "skill": "Classification",
      "tag": "new"
    },
    {
      "skill": "Ranking",
      "tag": "new"
    },
    {
      "skill": "Fraud Detection",
      "tag": "new"
    },
    {
      "skill": "Risk Modeling",
      "tag": "new"
    }
  ],
  "persistence": {
    "items": [
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Project Delivery and Coordination",
          "id": 366,
          "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
          "slug": "d_init_02",
          "source": "db"
        },
        "dimension_id": 366,
        "input_skill": "Machine Learning",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2672,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Machine Learning",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2672,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Analytical Programming Languages",
          "id": 82,
          "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
          "slug": "analytical-programming-languages",
          "source": "db"
        },
        "dimension_id": 82,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Analyst",
            "id": 20,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-analyst",
            "source": "db"
          },
          {
            "display_name": "Data Scientist",
            "id": 7,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-scientist",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Automation Scripting and CLI",
          "id": 48,
          "rationale": "Uses scripts and command-line tooling to execute repeatable Azure operations and reduce manual work. This is a practical cluster because the role frequently automates provisioning, checks, and remediation tasks.",
          "slug": "automation-scripting-and-cli",
          "source": "db"
        },
        "dimension_id": 48,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Azure Cloud Engineer",
            "id": 4,
            "rationale": null,
            "role_archetype": null,
            "slug": "azure-cloud-engineer",
            "source": "db"
          },
          {
            "display_name": "Cloud Engineer",
            "id": 18,
            "rationale": null,
            "role_archetype": null,
            "slug": "cloud-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Automation and Scripting for Operations",
          "id": 361,
          "rationale": "Scripts and lightweight automation used to execute repetitive virtualization tasks and enforce operational consistency. This is the practical glue that reduces manual host and VM administration.",
          "slug": "automation-and-scripting-for-operations",
          "source": "db"
        },
        "dimension_id": 361,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Virtualization Engineer",
            "id": 26,
            "rationale": null,
            "role_archetype": null,
            "slug": "virtualization-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Network Automation and Scripting",
          "id": 285,
          "rationale": "Covers scripts and automation used to configure, validate, and audit network devices and services. This cluster is coherent because repeatable network operations increasingly depend on programmatic changes and checks.",
          "slug": "network-automation-and-scripting",
          "source": "db"
        },
        "dimension_id": 285,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Network Engineer",
            "id": 21,
            "rationale": null,
            "role_archetype": null,
            "slug": "network-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Programming Languages for AI Workflows",
          "id": 261,
          "rationale": "Languages used to implement AI feature logic, orchestration, and response handling inside product code. This is the core coding surface for turning prompts and model calls into reliable application behavior.",
          "slug": "programming-languages-for-ai-workflows",
          "source": "db"
        },
        "dimension_id": 261,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "AI Engineer",
            "id": 12,
            "rationale": null,
            "role_archetype": null,
            "slug": "ai-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Programming Languages for Backend Systems",
          "id": 140,
          "rationale": "Languages used to implement server-side business logic, request handlers, workers, and service integrations. This is the core coding surface for backend feature delivery and maintenance.",
          "slug": "programming-languages-for-backend-systems",
          "source": "db"
        },
        "dimension_id": 140,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Backend Engineer",
            "id": 14,
            "rationale": null,
            "role_archetype": null,
            "slug": "backend-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Programming Languages for Data Work",
          "id": 67,
          "rationale": "Languages used to implement data pipelines, transformations, and operational utilities. This is the code layer for expressing extraction, parsing, validation, and orchestration logic in data engineering workflows.",
          "slug": "programming-languages-for-data-work",
          "source": "db"
        },
        "dimension_id": 67,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Engineer",
            "id": 6,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Programming Languages for ML Systems",
          "id": 113,
          "rationale": "Languages used to implement model integration code, inference services, and feature-processing logic. This is the core coding surface for turning trained models into product-facing software components.",
          "slug": "programming-languages-for-ml-systems",
          "source": "db"
        },
        "dimension_id": 113,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": true,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension saved",
        "role_dimension_saved": true,
        "roles_from_db": [
          {
            "display_name": "Machine Learning Engineer",
            "id": 10,
            "rationale": null,
            "role_archetype": null,
            "slug": "machine-learning-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Programming Languages for Security Work",
          "id": 328,
          "rationale": "Languages used to automate security tasks, write detection logic, and build analysis or remediation tooling. This is the core coding surface for a cybersecurity engineer across scripts, queries, and small utilities.",
          "slug": "programming-languages-for-security-work",
          "source": "db"
        },
        "dimension_id": 328,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Cybersecurity Engineer",
            "id": 9,
            "rationale": null,
            "role_archetype": null,
            "slug": "cybersecurity-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Programming Languages for Test Automation",
          "id": 193,
          "rationale": "Languages used to implement automated checks, helper utilities, and test harness code. This is the core coding surface for turning test ideas into maintainable automation.",
          "slug": "programming-languages-for-test-automation",
          "source": "db"
        },
        "dimension_id": 193,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Automation Tester",
            "id": 16,
            "rationale": null,
            "role_archetype": null,
            "slug": "automation-tester",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Security Automation and Scripting",
          "id": 258,
          "rationale": "Automating repeatable security checks, enrichment, and remediation workflows. This cluster is coherent because the role often needs lightweight automation to scale analysis and response.",
          "slug": "security-automation-and-scripting",
          "source": "db"
        },
        "dimension_id": 258,
        "input_skill": "Python",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Cybersecurity Engineer",
            "id": 9,
            "rationale": null,
            "role_archetype": null,
            "slug": "cybersecurity-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 393,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Applied Machine Learning Toolkits",
          "id": 94,
          "rationale": "Libraries and frameworks used to prototype and compare models quickly. This dimension captures the concrete tooling layer beneath modeling methods and evaluation.",
          "slug": "applied-machine-learning-toolkits",
          "source": "db"
        },
        "dimension_id": 94,
        "input_skill": "PyTorch",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Scientist",
            "id": 7,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-scientist",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 557,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Applied Machine Learning Toolkits",
          "id": 94,
          "rationale": "Libraries and frameworks used to prototype and compare models quickly. This dimension captures the concrete tooling layer beneath modeling methods and evaluation.",
          "slug": "applied-machine-learning-toolkits",
          "source": "db"
        },
        "dimension_id": 94,
        "input_skill": "TensorFlow",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Scientist",
            "id": 7,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-scientist",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 558,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Analytical Programming Languages",
          "id": 82,
          "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
          "slug": "analytical-programming-languages",
          "source": "db"
        },
        "dimension_id": 82,
        "input_skill": "pandas",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Analyst",
            "id": 20,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-analyst",
            "source": "db"
          },
          {
            "display_name": "Data Scientist",
            "id": 7,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-scientist",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 474,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Analytical Programming Languages",
          "id": 82,
          "rationale": "Languages used to clean, transform, analyze, and prototype models in notebooks and scripts. This is the core coding surface for expressing statistical logic and data manipulation in a reproducible way.",
          "slug": "analytical-programming-languages",
          "source": "db"
        },
        "dimension_id": 82,
        "input_skill": "NumPy",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Analyst",
            "id": 20,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-analyst",
            "source": "db"
          },
          {
            "display_name": "Data Scientist",
            "id": 7,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-scientist",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 475,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Project Delivery and Coordination",
          "id": 366,
          "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
          "slug": "d_init_02",
          "source": "db"
        },
        "dimension_id": 366,
        "input_skill": "OCR",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2653,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "OCR",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2653,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Experiment Design and Analysis",
          "id": 87,
          "rationale": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This cluster is central when the role needs to compare interventions, treatments, or product changes with rigor.",
          "slug": "experiment-design-and-analysis",
          "source": "db"
        },
        "dimension_id": 87,
        "input_skill": "A/B Testing",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Data Scientist",
            "id": 7,
            "rationale": null,
            "role_archetype": null,
            "slug": "data-scientist",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 506,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Cloud Service Integration Patterns",
          "id": 188,
          "rationale": "Covers how cloud services and workloads connect through APIs, events, shared services, and integration boundaries. This cluster is coherent because architects must define interaction patterns that preserve decoupling, security, and operability.",
          "slug": "cloud-service-integration-patterns",
          "source": "db"
        },
        "dimension_id": 188,
        "input_skill": "RAG",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "Cloud Architect",
            "id": 11,
            "rationale": null,
            "role_archetype": null,
            "slug": "cloud-architect",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 2659,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Context Management and Retrieval",
          "id": 264,
          "rationale": "Preparing, selecting, and packaging context for model calls so responses stay relevant and grounded. This is a distinct cluster because AI features often depend on what information is included, summarized, or retrieved at call time.",
          "slug": "context-management-and-retrieval",
          "source": "db"
        },
        "dimension_id": 264,
        "input_skill": "RAG",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "AI Engineer",
            "id": 12,
            "rationale": null,
            "role_archetype": null,
            "slug": "ai-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 2659,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Context Management and Retrieval",
          "id": 264,
          "rationale": "Preparing, selecting, and packaging context for model calls so responses stay relevant and grounded. This is a distinct cluster because AI features often depend on what information is included, summarized, or retrieved at call time.",
          "slug": "context-management-and-retrieval",
          "source": "db"
        },
        "dimension_id": 264,
        "input_skill": "Prompt Engineering",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [
          {
            "display_name": "AI Engineer",
            "id": 12,
            "rationale": null,
            "role_archetype": null,
            "slug": "ai-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 2683,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Feature Stores",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2700,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": null,
          "display_name": "Production Model Serving Architecture",
          "id": null,
          "rationale": "Patterns for hosting, routing, versioning, and scaling model inference in production systems, including online inference endpoints, batch scoring services, streaming inference, and serving topologies such as dedicated services, sidecars, gateways, or embedded serving layers. This skill covers request routing, autoscaling serving replicas, canary rollout for models, model version selection, and latency/throughput tradeoffs.",
          "slug": "d_merge_01",
          "source": "llm"
        },
        "dimension_id": 115,
        "input_skill": "Model Serving",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (reconciliation merge) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2701,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Model Serving Deployment and Runtime Packaging",
          "id": 52,
          "rationale": "Operational deployment of trained models into online, batch, or streaming serving environments, including packaging models and model servers into containers or managed inference runtimes, coordinating rollout, and handing off to inference systems. Covers serving frameworks and platforms such as TensorFlow Serving, TorchServe, Triton Inference Server, BentoML, KServe, and Seldon Core, plus container/runtime concerns like Docker images, GPU-enabled containers, base image selection, container entrypoints, runtime dependencies, and image scanning for model services.",
          "slug": "model-serving-deployment-and-runtime-packaging",
          "source": "db"
        },
        "dimension_id": 52,
        "input_skill": "Model Serving",
        "llm_role": null,
        "matched_chosen_role": true,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension saved",
        "role_dimension_saved": true,
        "roles_from_db": [
          {
            "display_name": "MLOps Engineer",
            "id": 5,
            "rationale": null,
            "role_archetype": null,
            "slug": "mlops-engineer",
            "source": "db"
          },
          {
            "display_name": "Machine Learning Engineer",
            "id": 10,
            "rationale": null,
            "role_archetype": null,
            "slug": "machine-learning-engineer",
            "source": "db"
          }
        ],
        "skill_dimension_saved": true,
        "skill_id": 2701,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Reinforcement Learning",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2702,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Graph Neural Networks",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2703,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Project Delivery and Coordination",
          "id": 366,
          "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
          "slug": "d_init_02",
          "source": "db"
        },
        "dimension_id": 366,
        "input_skill": "Graph Neural Networks",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2703,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": null,
          "display_name": "Experimental Design, Testing, and Causal Analysis",
          "id": null,
          "rationale": "Designing controlled experiments and analyzing results to estimate causal impact. This includes A/B tests, randomized experiments, hypothesis testing, metric selection, statistical significance, treatment/control design, power analysis, causal inference, and evaluating model or product variants through rigorous comparison. It covers the experimental method and interpretation of outcomes, not the mechanics of model training or data preparation.",
          "slug": "d_merge_01",
          "source": "llm"
        },
        "dimension_id": 87,
        "input_skill": "Fine-tuning",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (reconciliation merge) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2704,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Fine-tuning",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2704,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Classification",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2705,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Ranking",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2706,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Fraud Detection",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2707,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Project Delivery and Coordination",
          "id": 366,
          "rationale": "Coordination practices for organizing work, tracking progress, and aligning stakeholders across a delivery effort. Agile fits here when used as a team execution framework for managing scope, cadence, and collaboration.",
          "slug": "d_init_02",
          "source": "db"
        },
        "dimension_id": 366,
        "input_skill": "Fraud Detection",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2707,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": null,
          "display_name": "Experiment Design, Causal Inference, and Analysis",
          "id": null,
          "rationale": "Designing controlled experiments and analyzing outcomes to estimate causal impact. This includes A/B testing, randomized controlled trials, hypothesis testing, confidence intervals, causal inference, uplift analysis, backtesting, and treatment effect estimation. The focus is on rigorous causal measurement of interventions, treatments, or product changes, not on risk modeling, deployment, or reporting.",
          "slug": "d_merge_01",
          "source": "llm"
        },
        "dimension_id": 87,
        "input_skill": "Risk Modeling",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (reconciliation merge) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2708,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "well_known",
          "display_name": "Version Control Systems",
          "id": 365,
          "rationale": "Tools and workflows for tracking source changes, branching, merging, and collaborating on code history. Git belongs here because it is the canonical distributed version control system used to manage revisions and coordinate team development.",
          "slug": "d_init_01",
          "source": "db"
        },
        "dimension_id": 365,
        "input_skill": "Risk Modeling",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (library) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2708,
        "skill_tag": "in_db",
        "skipped_reason": null
      },
      {
        "chosen_role_id": 10,
        "dimension": {
          "difficulty_hint": "emerging",
          "display_name": "Model Serving Deployment and Runtime Packaging",
          "id": null,
          "rationale": "Operational deployment of trained models into online, batch, or streaming serving environments, including packaging model artifacts and model servers into containers or managed inference runtimes, managing runtime dependencies, defining deployment manifests and serving images, and coordinating rollout automation. Includes serving frameworks such as TensorFlow Serving, TorchServe, Triton, BentoML, KServe, and Seldon Core, plus Docker and GPU-enabled container concerns.",
          "slug": "d_merge_02",
          "source": "llm"
        },
        "dimension_id": 52,
        "input_skill": "Model Serving",
        "llm_role": null,
        "matched_chosen_role": false,
        "outcome_line": "New skill saved \u00b7 Existing dimension (reconciliation merge) \u00b7 Role\u2194dimension skipped (dimension not under chosen role)",
        "role_dimension_saved": false,
        "roles_from_db": [],
        "skill_dimension_saved": true,
        "skill_id": 2701,
        "skill_tag": "in_db",
        "skipped_reason": null
      }
    ],
    "new_skills_created": 9,
    "role_dimension_saved": 0,
    "skill_dimension_saved": 14,
    "skipped": 0
  },
  "planner_output": null,
  "run_id": "317b22e6-ddd9-4256-bb2b-552034ba2d8e"
}