# FastAPI · RAG · LLM agents · browser automation · 8+ yrs
Async ingestion pipeline over 200K scientific papers. FAISS + pgvector with query routing, hybrid semantic/keyword reranking. FastAPI service layer, measured retrieval quality vs ground truth.
Full-stack AI chatbot for real clients: FastAPI backend, Facebook Messenger API, WooCommerce webhooks, Playwright E2E test suite. Production deploys with Caddy + systemd.
Multi-turn evaluation harness: automated runner, JSON logging, 20+ failure cases, gate priming experiments, 3-arm A/B studies with reproducible CLI. Used to measure model degradation.
92-case adversarial test suite across 10 OWASP-style attack categories. Regex pre-filter + instruction hardening + output validation. Built for a live client LLM deployment at RemoteBlue.
# available for
contract work, staff augmentation,
and interesting Python projects.
# specialties
RAG pipelines · LLM agents
FastAPI backends · browser automation
evaluation frameworks · data tooling