AI-powered virtual assistant for satellite communications

Challenge

The satellite communications (Satcom) sector is increasingly complex and data-intensive, requiring fast, reliable support for technical verification and feasibility assessments. ESA commissioned this project to explore how Artificial Intelligence, specifically Large Language Models (LLMs), could be harnessed to create a Virtual Assistant (VA) tailored to Satcom needs. 

The challenge lies in: 

- Adapting open-source LLMs to the highly specialized Satcom domain. 
- Ensuring the VA performs effectively in real-world interactions.  
- Validating its ability to support operational workflows and technical queries.  

This required a rigorous evaluation framework to assess the VA’s relevance, accuracy, and usability. 

Applied expertise 

RINA played a pivotal role in shaping the technical direction of the project. Our contributions included: 

- Use case definition: identifying key Satcom applications and end-user needs. 
- Workflow mapping: structuring operational processes to ensure seamless VA integration. 
- LLM training support: preparing and evaluating question-answer sets to optimize the VA’s learning and response quality. 

PI School led the overall initiative, coordinating with ESA and RINA to ensure strategic alignment and innovation. The consortium worked collaboratively to build a prototype VA capable of understanding and responding to Satcom-specific queries with precision. 

Impact 

The project goal is to validate the prototype of the Virtual Assistant, demonstrating its potential to: 

- Streamline support for Satcom operations.  
- Enhance user experience through intelligent, context-aware responses. 
- Serve as a foundation for future AI-driven tools in space communications.  

This initiative sets the stage for broader adoption of AI in Satcom, with promising applications in system-level feasibility assessments, technical verification, and real-time user support.