LIVE — try it//Architecture
VEGA — RAG assistant
A résumé you can talk to. A full RAG pipeline, running in production on this site.
Pipeline // data flow
01 // Context
A PDF résumé just sits there: a recruiter or a client can't ask a precise question or check a skill in a few seconds. I wanted living proof of what I deliver — a complete RAG system in production, not a throwaway demo.
02 // The challenge
Make a profile you can query in plain language, by voice, with sourced answers, acceptable latency and controlled costs — all hosted in production, not running locally on my laptop.
03 // Result / status
Live on this site. Pipeline deployed on Vercel + Supabase: document ingestion → embeddings → vector search (pgvector) → generation (gpt-4o-mini) → speech synthesis. A daily token cap and per-IP rate limiting keep costs under control. You can try it right now.