- Developed and validated regulatory credit risk models for structured securities (ABS, RMBS, CLO) across US and non-US exposures, supporting loss forecasting and capital computation under CCAR, CECL, and IFRS 9.
- Built logistic regression and econometric models to forecast Default Rates, Prepayment Speeds, and Loss Given Default under macroeconomic stress scenarios.
Designed Decision
Tree benchmarks with bootstrapped confidence intervals for uncertainty quantification and K-Fold Cross-Validation backtesting to assess forecast accuracy across regimes.
- Owned end-to-end development of three ML challenger models (Logistic Regression, SVM, Random Forest) to benchmark the production XGBoost Ginnie Mae Prepayment classifier. Drove feature engineering via Information Gain, Importance Ranking, and Recursive Feature Elimination; the optimized Logistic Regression achieved superior AUC-ROC and recall and was adopted as the benchmark for its interpretability.
- Validated a multi-asset Black-Scholes Hull-White interest rate model for stochastic rates valuation adjustments on multicurrency FX exotics.
- Selected for a high-visibility cross-functional assignment to independently evaluate the Firmwide Deposits Trend Forecasting Model governing $2.4T exposure. Re-engineered the testing pipeline in Python for improved explainability; conducted regime-varying econometric analysis on alternative MeV specifications and quantified forecast uncertainty via bootstrapped confidence intervals.
- Led cross-functional monitoring of 11 credit risk models — defined test specifications with model developers and delivered documentation for senior stakeholders.
- Owned performance monitoring of 80+ risk models across CIO and Treasury. Automated the end-to-end pipeline and integrated LLM-powered analytics to improve efficiency and scalability.
- Evaluated an LLM-based model review tool under the firm's AI efficiency initiative; identified critical failure modes through systematic validation and delivered recommendations to senior leadership.
Data Scientist (Internship) at Societe Generale (2021-08 – 2021-09)
- Developed ensemble machine learning classification models on the GLBA Fund Analytics project to detect adverse, high-risk financial entities from large-scale, noisy financial datasets spanning multiple sources; achieved >90% precision and recall, supporting risk-based decision frameworks.
