AI-Powered Research Assistant System Using Retrieval-Augmented Generation
Shekhar Pandey1, Janvi Panchal2, Vishal Gupta3, Armaan Vala4, Sachin More5
1,2,3,4 Student, Department of Artificial Intelligence and Machine Learning
Universal College of Engineering, Mumbai, India
5Assistant Professor, Department of Artificial Intelligence and Machine Learning
Universal College of Engineering, Mumbai, India
Abstract—The rapid growth of scientific publications has made it increasingly difficult for researchers to efficiently discover and analyse relevant academic content. Traditional information retrieval systems focus primarily on keyword-based search, while modern AI systems emphasize text generation, often lacking reliable grounding in scholarly sources. To address these limitations, this paper proposes an AI-powered research assistant system that integrates semantic information retrieval techniques with Retrieval-Augmented Generation (RAG). The proposed system enables users to input natural language queries and retrieves relevant research papers using similarity-based retrieval methods. The retrieved content is then processed using generative AI models to produce concise summaries, extract key insights, and provide citation-aware responses. By combining retrieval mechanisms with generation models, the system reduces hallucination and improves the relevance and reliability of outputs. Furthermore, the system incorporates concepts from recommendation systems and natural language processing to enhance search efficiency and user interaction. Experimental observations indicate that the proposed approach improves research accessibility, reduces the time required for literature review, and provides a unified platform for academic knowledge discovery. This work demonstrates the potential of integrating retrieval and generation techniques for building intelligent research support systems.
Keywords— AI Research Assistant, Retrieval-Augmented Generation (RAG), Natural Language Processing, Information Retrieval, Academic Knowledge Discovery
