Problem
Questions about official university documents and long YouTube videos need answers grounded in the source, not guesses.
What I built
- Built RAG assistants to ingest official IUB PDFs and YouTube transcripts (via yt-dlp), storing vectors in PGVector.
- Orchestrated grounded answer generation with Llama 3.1 via Groq and MMR retrieval behind Streamlit interfaces.
Result
Two assistants that answer from official IUB PDFs and YouTube transcripts, using PGVector, MMR retrieval, and Llama 3.1 via Groq behind Streamlit interfaces.
Need something like this built?
Book a 30-minute intro call and tell me about the role or the project.
Next case study: Reusable GoHighLevel Real Estate CRM & Automation Snapshot