A Multi-Agent System Integrating Llm For Intelligent Athlete Assistance

This study presents a modular multi-agent system (MAS) integrated with large language models (LLMs) to support athletes in four domains: training and fitness, injury and rehabilitation, nutrition and mental health. Each specialised agent operates independently while collaborating to provide personalised, context-aware recommendations based on real-time athlete data. The research evaluates the suitability of different LLMs for each domain-specific agent using automated and human evaluation methods. A case study demonstrates the system's practical application and highlights its impact on athlete outcomes. The results demonstrate the system's ability to interpret user queries and deliver personalised, context-specific responses effectively. However, key limitations were identified, including communication issues between the mobile application and the backend, as well as constraints associated with deploying models locally. Nevertheless, the study showcases the potential of integrating LLMs into mobile applications to support athletes, emphasising the modularity of this approach and its capacity to deliver dynamic, personalised support.