At IndusInd Bank, I developed a behavioral credit-scoring model for a microfinance portfolio, training XGBoost on historical repayment records and selecting 18 high-signal predictors. The final model delivered a 34% KS and 28% Gini uplift over the incumbent, and its risk scorecard reduced annualized NPAs by ~₹115 Cr.
I also built an agentic AI HR assistant with a RAG pipeline and LangGraph workflow, and created SAS workflows that automated monthly processing across multiple credit bureaus. In a personal real-estate project, I developed an XGBoost price predictor that reduced MAE by 33% over the baseline.
