On Tuesday, October 20, Víctor Rodríguez and Alejandro Mosquera successfully defended their thesis at Universidad Tecnológica de Panamá (UTP). The work was developed through the SUCCESS project under the supervision of Aris Castillo (UTP) and Afshin Ameri (Mälardalen University, MDU).
Their thesis, “Implementation and Practical Evaluation of AI Agents in the UTP Smart Room Using the Model Context Protocol,” presents Jarvis, a local voice assistant for Smart Rooms. Jarvis is built on two AI agents: a router, which resolves most requests, and a thinker, which reasons through the more complex ones. Both connect to the room’s devices through the Model Context Protocol (MCP).
Rather than requiring exact commands, Jarvis interprets natural-language requests, greets people proactively as they enter the room, and supports open-microphone conversation without needing to be addressed by name. The thesis also proposes the harness, the engineering layer that defines the tools, the context and the model’s response, as the element that turns a language model into a room that acts.
The system was evaluated with 20 participants from outside the development team, measuring usability, agreement between participant and observer, the rate of unanswered turns, and response time. The results show the approach is viable as a proof of concept and a step toward smart spaces that understand people, not just their commands.
The examiners were UTP professors Kexy Rodríguez and Ernesto García. The defense was also attended by Radu Dobrin (MDU), SUCCESS project coordinator, and Sergio Gutiérrez (Universidad de Antioquia), a member of the SUCCESS project team.
Co-funded by the European Union.