Chat with Your Documents

Built during Infosys Springboard 5.0 internship

Developed an AI-powered tool to interact with multiple documents and retrieve insights from various document formats using Retrieval-Augmented Generation (RAG) and NLP techniques — with text chunking, embeddings stored as vectors, a semantic retriever, and conversational memory for follow-up questions.

• Main Challenge in the project: Getting the retriever to pull the right chunks was the biggest challenge — every document format breaks text differently, so tuning chunk size and overlap, picking embeddings that capture meaning accurately, and keeping conversational memory coherent across follow-ups without the bot hallucinating took the most work.