classimbalance

End-to-end ML pipeline for stroke risk prediction with XGBoost, FastAPI API, and Streamlit dashboard — designed to handle severe class imbalance and maximize recall.

WeLT-ASpERT Trainer

Unofficial pytorch implementation on DAR-BN

A experimental project that delves in dealing with CI dataset using SMOTENN+GAN with dvc experimentation.

WeLT TablERT-CNN Trainer

ML pipeline predicting loan default risk — imbalance-aware training, cost-based thresholding, and SHAP explainability on 32K+ records.

WeLT-SpERT Model