At ATL Leasing, I built an end-to-end PD scoring model in Python using more than 17,000 leasing contracts. It achieved an AUC of 0.93 against the expert baseline of 0.58, in line with Basel II/III and IFRS 9.
I also cleaned and consolidated contract data, engineered more than 30 risk features, and validated models using cross-validation, Brier Score, and PSI. I used SHAP to explain model drivers and flagged concentration risk with a shadow-scoring recommendation.

