FairLens — A Forensic Audit of Algorithmic Bias in US Mortgage Lending
shauryaditiyas
About
A complete fairness audit of an XGBoost mortgage approval model trained on 50,000 HMDA applications. Identifies federal lending law violations, detects ZIP code as a proxy for race via SHAP, and applies Fairlearn's ThresholdOptimizer to restore compliance for all four minority groups. Submitted to ZerveHack 2026.
