Model Governance Framework

Comprehensive AI governance framework for regulated industries with bias detection, fairness metrics, and model documentation.

Business Impact

• FDA-ready model documentation with automated model cards
• Bias detection across protected demographic attributes
• Explainability with SHAP and LIME integration
• Zero compliance violations across $51M+ portfolio

Key Features

Fairness and Bias Detection

• Demographic Parity: Equal positive rates across groups
• Equalized Odds: Equal TPR and FPR across groups
• Disparate Impact: 4/5ths rule compliance
• Intersectional Analysis: Multi-attribute fairness

Explainability

• SHAP value computation
• LIME explanations
• Feature importance analysis
• Decision boundary visualization

Model Documentation

• Automated model cards (Google format)
• Intended use documentation
• Limitation and risk disclosure
• Healthcare-specific compliance notes

Quick Start

from src.fairness.bias_detector import BiasDetector, ProtectedAttribute

detector = BiasDetector()

report = detector.analyze(
    model_id="diabetes-classifier",
    model_version="1.0.0",
    predictions=[...],
    actuals=[...],
    protected_attributes={
        ProtectedAttribute.GENDER: ["male", "female", ...],
        ProtectedAttribute.AGE_GROUP: ["18-40", "41-65", "65+", ...],
    },
)

print(f"Overall Status: {report.overall_status}")
for violation in report.violations:
    print(f"  - {violation.metric}: {violation.description}")

Project Structure

model-governance-framework/
├── src/
│   ├── fairness/
│   │   └── bias_detector.py     # Fairness metrics
│   ├── explainability/
│   │   └── explainer.py         # SHAP/LIME integration
│   ├── cards/
│   │   └── model_card.py        # Documentation generation
│   └── documentation/
│       └── report_generator.py  # Compliance reports
├── tests/
└── templates/

Author

Christopher Mangun - github.com/cmangun