University support systems are under increasing pressure to handle high volumes of student queries accurately and at scale. Traditional rule-based chatbots are rigid and brittle, while large language model (LLM)-based systems, though fluent, are prone to hallucination. This paper presents UNI ASSISTAI, a Retrieval- Augmented Generation (RAG)-based intelligent university assistant that grounds every generated response in verified institutional knowledge. The system integrates a semantic vector retrieval pipeline with a GPT-based generative model, and extends it with multimodal input capabilities—supporting text, voice (via ASR), and image (via OCR) queries. The backend is served through a FastAPI interface, and the frontend is implemented in React with TypeScript and Tailwind CSS. Experimental evaluation on a curated university FAQ and policy corpus yields a Precision of 0.87, Recall of 0.84, and F1-score of 0.85, outperforming both rule-based and vanilla LLM baselines. This work demonstrates that domain-specific RAG architectures offer a scalable, reliable path to academic AI assistants.
Kumar et al. (Thu,) studied this question.