Enhancing Large Language Models with a Hybrid Retrieval-Augmented Generation System: A Comparative Analysis

Abstract: Organizations are increasingly challenged with maintaining data privacy when utilizing cloud-based AI services for natural language processing tasks. This project aims to tackle these issues by creating a secure, on-premise system that employs Retrieval-Augmented Generation (RAG) to allow organizations to query their internal data safely. Our solution combines transformer models with document retrieval techniques to produce contextually relevant answers to user queries, all while ensuring that data stays within the organization's local network. This method assists organizations in adhering to strict data privacy regulations such as GDPR and HIPAA. Built entirely in Python, our system is designed for scalability, flexibility, and seamless integration with existing infrastructure.