Dave Ebbelaar builds a full-stack RAG app in a 4-hour tutorial
AI engineer Dave Ebbelaar walks through building a full-stack Document Copilot from scratch — a RAG application over SEC filings — using FastAPI, React, Supabase, pgvector, and Railway, covering the complete architecture from ingestion to deployment in roughly 4 hours.
Score breakdown
The tutorial is notable for covering the complete full-stack architecture of a production-style RAG application — frontend, backend, database, ingestion pipeline, and deployment — in a single end-to-end walkthrough, which Ebbelaar describes as rarely seen on YouTube.
- 01The project is a Document Copilot — a RAG app for querying SEC filings with grounded answers, citations, and chat history.
- 02Stack includes FastAPI, Supabase Postgres, pgvector, React, TypeScript, Tailwind, shadcn/ui, and Railway.
- 03Tutorial covers frontend auth and chat, backend, ingestion, chunking, embeddings, hybrid search, retrieval, and deployment.
Dave Ebbelaar, AI engineer and founder of Datalumina, released a roughly 4-hour end-to-end tutorial building a "Document Copilot" — a RAG application that allows users to ask questions over SEC filings and receive grounded answers with citations and chat history. The project uses FastAPI for the backend, Supabase Postgres with pgvector for the database and vector storage, React with TypeScript, Tailwind, and shadcn/ui for the frontend, and Railway for deployment. The tutorial also covers document ingestion, chunking, embeddings, and hybrid search as part of the full retrieval pipeline.
Ebbelaar frames the video as a real-world AI engineering walkthrough rather than a beginner-friendly course, noting he moves at his natural pace and uses AI agents throughout the build process.
Ebbelaar frames the video as a real-world AI engineering walkthrough rather than a beginner-friendly course, noting he moves at his natural pace and uses AI agents throughout the build process. A GitHub repository is provided with both a starting-point branch and a completed development branch, so viewers can code along or reference the final result. He describes the video as condensing approximately 40 hours of learning into the 4-hour runtime, and directs viewers to his broader YouTube channel for deeper dives into individual topics such as hybrid search, agentic search, and CI/CD pipelines.
Key facts
- 01The project is a Document Copilot — a RAG app for querying SEC filings with grounded answers, citations, and chat history.
- 02Stack includes FastAPI, Supabase Postgres, pgvector, React, TypeScript, Tailwind, shadcn/ui, and Railway.
- 03Tutorial covers frontend auth and chat, backend, ingestion, chunking, embeddings, hybrid search, retrieval, and deployment.
- 04A GitHub repository is provided with a starting-point (main) branch and a completed (development) branch.
- 05Ebbelaar describes the video as condensing approximately 40 hours of learning into the 4-hour runtime.
- 06The tutorial is aimed at developers seeking production-style AI engineering experience, not beginners looking for step-by-step guidance.
- 07Ebbelaar is the founder of Datalumina and uses AI agents during the build process.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 9, 2026 · 17:05 UTC. How this works →