All projects

03GenAI / LLM Prototype / RAG

PdfInsight

Ask questions of long PDFs and get grounded answers, using RAG with a local LLM.

The problem

Getting a specific answer out of a long PDF is slow, and plain keyword search is a blunt tool for it.

What I did

  • RAG over long PDFs with Ollama as the LLM backend.
  • Semantic search (embeddings + FAISS/Chroma) improved query efficiency by 65% over keyword search.
  • Modular FastAPI service, reusable in other AI workflows.

Tools & focus

  • Python
  • FastAPI
  • LangChain
  • Ollama
  • FAISS