At CVS Health, I owned a GPT-4 assistant that uses RAG over clinical and policy documents, cutting lookup time by 80% for 70 care-team analysts. I also built an LLM evaluation pipeline that raised response accuracy from 88% to 95% and was adopted by two teams.
Previously, I led a team of five at Chubb building claims risk-scoring and fraud-detection models. At Amdocs, I created a reusable ingestion framework that cut new-source onboarding effort by 40%, and at IBM (Client: MAPFRE Insurance), I built ETL pipelines for predictive modeling and analytics.
