First prototype · in build · months 1 to 3
Credit-model audit toolkit
Tests a lending or risk model for bias, stability and explainability. Aimed at banks, insurers and supervisors. Open-source code and a public-data demo will be linked here when they exist. There is no GitHub repository to pretend about yet.
Scope
A model card: task, population, features, decision threshold, and the claim under test.
Bias
Whether approval or score differs across groups in that book. We will publish the method with the first demo, not a mock chart.
Stability
Performance across time windows and population slices. Drawn as a heatmap in the lab palette.
Explainability
Feature contributions a credit committee can read. We will publish the method with the first demo.