Applied AI Engineering
Full-Stack AI Engineer
I build production-ready AI SaaS products end to end — agent orchestration, retrieval pipelines, async job processing, and the auth, billing-adjacent infra, and deployment plumbing that turns a working model into a system someone can actually rely on.
My focus is applied AI engineering — taking a model or a prompt chain and wrapping it in the infrastructure that makes it a real product: job queues for long-running work, usage quotas that keep API spend bounded, auth flows that actually verify an email, and a deployment that survives sharing a VPS with three other projects.
The three systems below are all live: a hybrid-RAG customer support agent, an enterprise knowledge assistant, and a video-to-shorts pipeline that runs real transcription and highlight-detection models against uploaded footage.
Converts long-form video into ~1-minute short clips for YouTube Shorts, TikTok, and Reels. Runs real transcription and Claude-based highlight detection as an async job pipeline, with burned-in captions and per-account usage quotas to keep processing costs bounded.
Agentic SaaS support system built on a hybrid retrieval strategy, with confidence-based web fallback when internal knowledge doesn't cover a question, safety guardrails on responses, and n8n-driven email automation for escalations.
Enterprise-style retrieval assistant for querying internal documentation, with a Streamlit frontend for quick internal rollout and a FastAPI backend handling embedding, retrieval, and citation-grounded answers.