RAG-powered portfolio assistant for technical documentation
A fully local RAG system that answers questions about technical portfolios, codebases, and documentation.
Business Challenge
Teams spend hours manually searching through documentation, code, and project files to answer questions.
Solution Delivered
Built an end-to-end RAG pipeline that ingests 291 files (code, markdown, Jupyter notebooks, PDFs), chunks them semantically, embeds with BGE, indexes with FAISS, and generates answers via local LLM (Ollama). All runs locally with no cloud dependencies.
Outcomes Achieved
Reduced document retrieval time from hours to <2 seconds per query Eliminated reliance on cloud APIs — full local control Enabled instant access to technical knowledge across 8,361 chunks and 291 files
Measurable Business Outcome
Reduced document retrieval time from hours of manual search to <2 seconds per query across 8,361 semantic chunks.