MLOps.community 9 Commandments for Building AI Agents
91 snips
Aug 1, 2025 
In this insightful discussion, Dmitri Jarnikov, Senior Director of Data Science at Prosus Group, and Paul van der Boor, VP AI at Prosus Group, delve into the complexities of building effective AI agents. They explore the importance of memory and the React cycle for continuous improvement. The conversation touches on balancing human involvement with AI decision-making, budgeting for operations, and the need for intuitive interfaces. Their practical approach offers a fresh perspective on navigating the challenges of AI development.
AI Snips
Chapters
Transcript
Episode notes
Expect Bottlenecks, Build or Buy Wisely
- Technology fast evolution means agents will always face bottlenecks needing custom solutions.
- Building solutions internally is often faster and cheaper than waiting for perfect external tools.
Empower Users with No-Code
- Provide non-engineers with no-code tools to build and customize AI agents.
- Enable users to integrate agents with common enterprise systems easily to solve their specific tasks.
Agents Learn From Experience
- Agents learn from their experiences to improve task success and efficiency.
- Recording successful and failed task paths creates an inherent memory to fine-tune reasoning models.