Enhancing large language models with a hybrid retrieval augmented generation system: A comparative analysis

Belhe, S. A. and Barse, Parth and Chingunde, Dheeraj and Katkar, Rutuja and Koul, Vansh (2025) Enhancing large language models with a hybrid retrieval augmented generation system: A comparative analysis. International Journal of Science and Research Archive, 15 (1). pp. 1607-1612. ISSN 2582-8185

With the increasing reliance on cloud-based AI services for NLP tasks, organizations are facing significant challenges in ensuring the privacy and security of their internal data. Stringent data privacy regulations like GDPR and HIPAA require organizations to safeguard sensitive information and prevent it from leaving their local infrastructure. This project proposes a solution to address these concerns by leveraging Retrieval-Augmented Generation (RAG) techniques, which combine transformer-based language models with document retrieval systems to generate accurate, contextually relevant responses while ensuring data remains within the organization’s local environment. By integrating an on-premise system for document retrieval and response generation, we ensure that sensitive information is never exposed to external cloud servers, helping organizations comply with privacy regulations. The entire system is implemented in Python, designed to be scalable, flexible, and seamlessly integrated into existing infrastructure, making it a practical solution for organizations seeking to utilize advanced AI capabilities without compromising data security. This approach not only enhances privacy but also enables organizations to harness the power of AI-driven NLP tasks safely and efficiently.