AIMP 15: Arnaud Jean on Building AI Agents, Games and Real-World Apps

What happens when an AI agent becomes the Game Master in a role-playing game? In episode 15 of AI Master Podcast, I talk with Arnaud Jean, Developer Advocate at AWS, about building AI agents through a project that is both useful and a lot of fun: an AI Game Master. We explore how agents can roll dice, check game rules, create characters and work together to tell a story. Along the way, Arnaud explains the ideas behind tools, MCP, A2A and RAG, and why learning these concepts matters far beyond games. You will hear about: • What makes an AI agent different from a chat with an LLM. • When one agent is enough, and when specialized agents can help. • The orchestrator pattern and the trade-offs between control and autonomy. • Building with Strands Agents and connecting tools through MCP. • Using a knowledge base to help an agent check the rules. • Starting locally and exploring Amazon Bedrock and AgentCore. • Testing agent behavior with traces, OpenTelemetry and MCP Inspector. • Why strong software fundamentals still matter as AI tools change. We also talk about board games, family and the joy of learning by building something a little crazy. Arnaud, thank you for such a fun and thoughtful conversation! AI Poland: https://ai-poland.pl More podcasts, people and ideas: https://kalbarczyk.dev Host: Dariusz Kalbarczyk Guest: Arnaud Jean Language: English #AIMasterPodcast #AgenticAI #StrandsAgents