Enhancing Career Counseling for Baccalaureate Students in Morocco through Retrieval-Augmented Generation Technology: A RAG-Based Chatbot

Abstract

Career guidance is important for helping students make informed choices about their studies and future careers. Yet many students lack access to resources that match their personal needs, leaving them uncertain about their options. This paper presents a chatbot built with Retrieval-Augmented Generation (RAG) to provide tailored career advice for baccalaureate students in a Moroccan high school. We compared multiple embedding models (OpenAI text embedding, Bidirectional Encoder Representations from Transformers (BERT), and Mistral) and large language models (GPT, LLaMA, and Mistral) within the same retrieval–generation framework. The evaluation showed that while all models achieved strong performance, the combination of OpenAI embeddings with Elasticsearch for retrieval and GPT for generation produced the most consistent and contextually appropriate responses. These findings show that RAG-based systems can support students with timely and personalized career guidance that connects their academic choices to available career options.