Built a custom Q&A chatbot for KGISL using a local language model to provide fast, context aware responses without relying on external APIs. Implemented a Retrieval Augmented Generation (RAG) pipeline to extract and index text from internal PDF documents, enabling accurate answers grounded in company knowledge. Designed a custom web interface for employees to query documents in natural language. Optimized embedding, chunking, and similarity search to improve response relevance and latency.