At Wells Fargo, I build and deploy full-stack AI platforms using React, FastAPI, and async Python, serving 400+ concurrent users during sprint-closing days.
I've shipped LLM-powered Sprint Goals Generator and Sprint Report Creator features, using few-shot prompting, Jira GreenHopper APIs, guardrails, and failure-recovery logging for reliable structured output. I also designed file-ingestion automation that reduced end-to-end latency by 40% and consolidated 28+ SQL workflows into a unified pipeline, reducing maintenance by 95%.
My work spans RAG, prompt caching, LLM routing with Gemini and FAISS, data reconciliation, CI/CD, and production deployment. At Code 4 Gov Tech, I built a Python RAG pipeline for large document corpora and improved retrieval accuracy by 15% through embedding evaluation frameworks.
Outside of work, I build real-time collaborative systems and production platforms, including an offline-first Kanban board with FastAPI, WebSockets, Kafka, and CRDTs, and a physiotherapy e-commerce platform deployed on GCP with Docker and GitHub Actions. I've also contributed to DragonflyDB, a Redis-compatible in-memory datastore.
