← All work
FLAGSHIP · MLFlagship ML case study
CreditSense AI
An explainable credit-risk scoring system — built so every decision comes with the reasons behind it, not just a number.
Role · ML engineer
PythonXGBoostSHAPFastAPIPostgreSQL
The idea
Credit models are often black boxes: a score comes out, but nobody can say why. CreditSense pairs gradient-boosted scoring with per-decision explanations, so each risk score arrives with the factors that drove it.
Approach
Explainability is a first-class requirement here, not an afterthought bolted on at the end.
- XGBoost for the core gradient-boosted risk model.
- SHAP attributions to explain each individual prediction — which features pushed a score up or down, and by how much.
- A FastAPI inference layer exposing the model as a service, backed by PostgreSQL.
Work in progress
The full technical write-up — data pipeline, evaluation methodology, and SHAP visualizations — is in progress. No performance metrics are quoted until they've been measured and confirmed.