Portfolio
Quantitative and data analysis work in stress testing, credit risk, machine learning, and SQL. Code available on GitHub.
Bayesian VAR over eleven US macro-financial variables that forecasts the economy and generates DFAST-style adverse scenarios. Sets severity with Growth-at-Risk and decomposes the Fed's severely adverse scenario into its macro core and its supervisory overlay. Includes a live interactive dashboard, and feeds the bank stress test below.
Reduced-form macro stress test of credit risk across ten large U.S. banks using a panel fixed-effects satellite model on FDIC call reports and FRED macro history. Maps the 2025 Fed severely adverse scenario to projected NPL paths and capital ratios, with a full validation suite and out-of-sample backtest.
Logistic regression on 1,000 German credit applicants with fairness analysis showing gender disparity at a single threshold and optimized group-specific cutoffs to equalize approval rates
Predicting personal loan interest rates across 15,000 borrowers using regression trees, best subset linear regression, and LASSO, with model comparison and residual analysis
Classification tree in R predicting startup outcomes across 1,290 companies using funding, team, and investor features, with pruning and centroid-based difficulty analysis
Classifying 3,000 CFPB consumer complaints using OpenAI embeddings and decision trees. PCA reduces error from 12.2% to 3.3%
AI-powered financial risk management dashboard for intelligent risk monitoring and analysis