At Nike, I designed production personalization, recommendation, and ranking systems, and built LLM workflows for retrieval, summarization, classification, and structured generation. I also developed repeatable evaluation pipelines that check quality and regressions against established baselines.
At Exelon, I reduced load-forecasting error from 12.7% to 6.3% through feature engineering, model tuning, and dataset validation. Earlier at Optum, I built Python services for chronic-disease risk scoring and supported predictive healthcare workflows using clinical data.
