RAG Solo — AI Engineer Year 2025
RAG-Powered Document Assistant
An assistant that answers questions in natural language over corporate PDFs and knowledge bases, always citing its sources.
Problem
Finding the right information across scattered documents took minutes, and the sources behind each answer were usually unclear.
Approach
- Chunking and embedding strategies were split by document type.
- Hybrid search (BM25 + vectors) and reranking improved context quality.
- Every answer now cites its sources, making the hallucination rate measurable.
Have a similar project?
Get in touch