T.Rushita Sree, S.Jaya Sri 03 06 The exponential growth of digital content across domains such as research legal and corporate knowledge bases has created an urgent need for systems capable of accurate contextaware document retrieval Traditional keywordbased search engines are often insufficient as they cannot capture semantic nuances understand context or synthesize information from multiple documents This paper presents a Document Retrieval System using RetrievalAugmented Generation RAG which combines retrievalbased embeddings with transformerbased generative models to provide highquality responses to user queries The proposed system first retrieves the most relevant documents from a large corpus using dense vector representations and semantic search Subsequently a generative language model synthesizes coherent contextrich answers by integrating information from these documents Experimental results demonstrate that the RAGbased system significantly outperforms traditional retrieval models in precision recall and user satisfaction metrics Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026