Minerva Lab
Practitioner-focused publications on fairness testing, governance frameworks, and bias audit methodology. Everything here is built from real analysis — we test methodology on real data, document what works, and share what we learn.
A build-once compliance guide for financial institutions navigating multiple AI governance frameworks. Artifact-first mapping showing what you need to produce, where frameworks overlap, and where they diverge — with a minimum viable evidence system and RACI by lifecycle stage.
Download DOCXThe 12 principles in plain English, a testing recipe with data requirements, red flags, and worked examples. Includes a 30/60/90-day implementation plan and stakeholder FAQ.
Download DOCX24-check assessment covering Principles 1–4 with red flag tables, evidence requirements, and a summary scorecard. Designed for model risk teams running their first FEAT assessment.
Download PDFAdverse impact analysis of 307,000 consumer loans from Home Credit Group's Southeast Asian portfolio — the same dataset used in the MAS Veritas Consortium case studies. Examines gender, age, and family status disparities using a simulated risk model with intersectional analysis.
Download DOCXIndependent adverse impact analysis of 50,000 mortgage applications. Demonstrates our methodology on real public data — including intersectional analysis that revealed disparities invisible in single-attribute testing.
Download DOCX