About This Project
Designed and implemented an end-to-end Retrieval-Augmented Generation (RAG) system that enables grounded, evidence-backed question answering over unstructured documents. Built a multi-stage pipeline including document ingestion, chunking, embedding generation, and semantic retrieval using ChromaDB. Integrated a multi-query retrieval strategy leveraging chat history with deduplication and dynamic ranking of chunks. Developed a FastAPI backend with a modular architecture and added real-time voice interaction using local Whisper for speech-to-text.
Key Features
- Multi-query retrieval using conversation history
- Deduplication of retrieved chunks across queries
- Dynamic top-k retrieval and ranking
- Evidence-backed answer generation using RAG
- Voice + text interface with local speech-to-text (Whisper)
- Modular pipeline (ingestion, indexing, retrieval, generation)
