
Session 21: Multi Agent Challenges
Presenters
Multi Agent Challenges
The event "Multi Agent System: From Challenges to Solutions" explored the concept of agents and multi-agent systems, highlighting their roles, collaboration methods (fully or partially connected), and practical applications, such as in software companies where specialized agents (e.g., Supervisor, Researcher, Developer) work together. Key challenges included managing too many tools, context window limitations, and handling large outputs. Solutions proposed leveraging multi-agent systems to break tasks into sub-tasks, summarize outputs, and augment queries, with examples like Pattern.global. Both multi-agent and single-agent systems were discussed, emphasizing their strengths in tool handling and response efficiency. The event also promoted an AI Talks Telegram group for further engagement.