I’ve built credit risk and predictive modeling systems at Kotak Mahindra Bank, including a dual-model credit card framework with short- and long-term risk signals, 8,400+ engineered features, and regulatory-aligned interpretability using IV and SHAP. I’ve taken models into PA/PQ decisioning workflows so LT/ST scores jointly inform approval/decline, swap-in/swap-out decisions, and dynamic line-of-credit capping.
Before that, at Kinara Capital I developed payment reminder and bounce collection models with XGBoost and Optuna, plus daily sales prediction using AWS Glue automation and business-facing visualization. Earlier, at Standard Chartered GBS, I worked on credit card default rate prediction with attribute engineering and hyperparameter tuning to improve ROC-AUC, and I’ve also built data tooling like a Java-based sync solution and an MVP dashboard workflow.
