A Multi-Agent Retrieval-Augmented Chatbot for Academic and Career Guidance

Abstract

The rapid growth of Artificial Intelligence (AI) and Natural Language Processing (NLP) has significantly improved student support services in higher education, particularly in academic advising and career counseling. However, traditional human-centered support systems often struggle to scale with increasing student populations, leading to delays and inconsistent guidance. This study proposes a dual-agent conversational chatbot that provides academic and career guidance using a Retrieval-Augmented Generation (RAG) framework. The system consists of two specialized agents—an Academic Agent and a Career Agent— both trained on domain-specific datasets to reduce semantic ambiguity and improve response accuracy. Sentence-BERT (SBERT) is used to generate semantic embeddings, while Facebook AI Similarity Search (FAISS) enables efficient semantic retrieval of relevant responses. A threshold-based response validation mechanism is integrated to prevent low-confidence answers and reduce misinformation. Experimental results show that the Academic Agent demonstrates stable training accuracy, while the Career Agent achieves strong retrieval performance validated through confusion matrix and Top-K accuracy analysis. The findings indicate that combining domain-separated agents with semantic retrieval and RAG architecture significantly improves response relevance, reliability, and transparency compared to general-purpose chatbot systems.