Guia UTFPR-CM: um agente de inteligência artificial para apoiar calouros na adaptação ao contexto do campus

Abstract

Integrating incoming students into the university environment poses a significant challenge for higher education institutions, particularly regarding efficient access to information on academic procedures, student services, and practical aspects of university life. This work presents an intelligent conversational agent's development and implementation process based on the Retrieval-Augmented Generation (RAG) architecture, specifically designed to support first-year students at the Federal University of Technology – Paraná (UTFPR), Campo Mourão campus. The system's primary knowledge base is the institutional "Student Guide", along with other documents and public notices provided by the university on its portal, transforming these static materials into a dynamic and interactive interface capable of responding to natural language queries. The methodology involved systematic analysis and content structuring, creating vector embeddings optimized for Brazilian Portuguese, implementing an RAG pipeline using Google Cloud Vertex AI technologies, and developing a simple web interface. The intended impact includes reducing in-person inquiries and decreasing the average response time to common questions. Comparative analysis with traditional methods demonstrated significant advantages in availability, speed, personalization, and scope. This work advances knowledge in applying RAG technologies in specific educational contexts, offering a replicable methodology for other higher education institutions.