Case study · 2026
NeuraFolio
Chat with a CV + GitHub knowledge base — RAG, citations, hybrid search, multi-provider fallback.

- Hybrid retrieval: pgvector similarity + full-text search over ingested CV and repos.
- SSE chat with source citations and follow-up suggestions grounded in retrieved chunks.
- Groq primary with Gemini/Cerebras fallback; NestJS API on Render + Next.js on Vercel.
Problem
Most “AI portfolio” demos are a thin chat wrapper over a prompt. The hard part is ingesting a CV and public repos into a searchable knowledge base, retrieving the right chunks, and answering with citations that stay honest when the model fails over.
Approach
NeuraFolio ingests a resume PDF and GitHub repos into Neon Postgres (pgvector + FTS), serves RAG chat from a NestJS backend with hybrid retrieval and optional tool agent, and streams answers to a Next.js UI with source filters (All / CV / GitHub) and temporary visitor CV upload.
Outcome
A live product you can open and chat with — creator CV + GitHub by default, upload-your-own CV for a session, and a clear split between frontend (Vercel) and API/RAG (Render) on a free-tier path.