The aim of this Master’s thesis is to design and implement an educational assistance system integrated with a social robot. The main objective is to provide students with a tool that supports their studies by providing coherent and reliable answers to questions, based on specific academic documents. Teaching assistants based on artificial intelligence have been shown to be highly beneficial, as they can adapt to each student’s learning pace and provide high-quality information at any time. Nevertheless, Large Language Models (LLMs), despite their considerable capacity for natural language processing, are subject to limitations such as reduced traceability and the generation of hallucinations. This work uses a Retrieval Augmented Generation (RAG) system, combining the generative capacity of an LLM with a verified document retrieval engine to mitigate hallucinations and allow greater transparency of sources. This work employs a modular and scalable architecture that utilises Docker and Docker Compose. It incorporates an Ollama model for text generation and combines a vector database with Chroma for information retrieval. It also includes an API that connects these functionalities with a web interface and a social robot. An innovative feature is the integration of a social robot, which can communicate responses through voice, facial expressions and movements. This generates a more immersive and intimate interaction that can adapt to the emotions conveyed during conversation with the user. In conclusion, this work presents an educational assistant with a hybrid architecture that combines the generative capacity of an LLM with a system for retrieving verified academic documents. This is all integrated with a social robot that serves as the system’s physical interface. Overall, this work is part of an UPM Educational Innovation Project and offers a viable, reliable, interactive learning tool that can be adapted to students’ needs.