Chain of Thought Mastering Multi-Agent Systems | MongoDB’s Mikiko Chandrasekhar
Jul 23, 2025
Mikiko Chandrasekhar, a Staff Developer Advocate at MongoDB, brings her expertise in the data-to-AI pipeline to the discussion on mastering multi-agent systems. She highlights the importance of reliability in AI agents, advocating for treating them as software products. The conversation delves into MongoDB’s role in supporting generative applications and discusses the need for robust debugging tools and systematic evaluation for agent performance. Mikiko also emphasizes the significance of balancing qualitative insights with quantitative metrics for effective AI development.
AI Snips
Chapters
Transcript
Episode notes
Agent Reliability Is Paramount
- Mastering agent reliability is crucial for AI systems to function effectively in production.
- Observation, evaluation, and guardrails are essential components of reliable agentic systems.
MongoDB as Agent Memory Store
- MongoDB acts as a flexible, polymorphic data store highly suitable for supporting AI agent systems.
- Beyond storage, MongoDB supports indexing, memory stores, and debugging tools to enhance agent application performance.
Real Production Agent Successes
- Despite skepticism in tech circles, many startups and large companies are building production-scale AI agents.
- MongoDB's recent acquisition of Voyage integrates embedding and re-ranking models to enhance agent performance.