At NPCI, I led a four-member cross-functional team to develop and productionize a profile-level fraud model for UPI person-to-person transactions. My XGBoost classifier increased UPI P2P fraud capture by 1% in value terms.
I use exploratory analysis and feature engineering to identify fraud patterns across transaction data, including payment escalation, transaction velocity, and geo-location signals. I also defined a modelling population covering roughly 40% of monthly UPI P2P fraud by value.
Previously at NSE, I co-authored Market Pulse and thematic macroeconomic reports, assessed stock liquidity using tick-by-tick order data, and redesigned a capacity forecasting framework in line with SEBI directives.
At CAFRAL, RBI, I researched corporate credit risk and climate-related macroeconomic shocks using OLS, panel, and fixed-effect regression models. I bring economics, financial-market research, and practical predictive modelling together to solve data-driven problems.
