At Autism360, I built and expanded a conversational assistant for families and clinicians on a healthcare SaaS platform. I worked with product managers and engineers, and personalized responses using each family's history and care plan.
I designed and refined versioned prompts, then tested them against golden sets and clinician review. I also built an evaluation suite with regression checks and LLM-as-judge graders calibrated against human ratings.
I wrote the Python services and FastAPI endpoints behind the assistant, with PostgreSQL and pgvector supporting its workflows. Response latency fell 55%, and incident resolution dropped from 8 hours to under 1 hour.
At Sephora, I shipped user-facing AI for retail, including a support assistant and a personalized recommendation engine. Earlier, I worked on ML platforms, model optimization, and data pipelines at Blue People and Ensitech; I also build projects including a personal finance AI coach and an agent evaluation harness.

