At Ally Bank, I redesigned card-fraud scoring with an ensemble of XGBoost, PyTorch, and TensorFlow models, reducing false positives by ~25% and raising the catch rate by ~18%. I also shipped conversational RAG agents that drafted citation-backed case summaries, helping cut fraud investigation time from days to minutes.
At CubeSmart Self Storage, I led a machine-learning personalization program that lifted customer conversion by ~18% and reduced churn by ~8%. Earlier, I worked across production ML, data pipelines, and cloud migrations at Teradata, Pennsylvania State University, Brillio Technologies, and Accenture.
