Self-Improving Agents and Agent Evaluation With Arize and Databricks MLflow

Explore the critical aspects of building resilient, self-improving autonomous agents in this 28-minute conference talk. Learn how to develop agents capable of evaluating their own performance and continuously improving through iterative feedback loops, while addressing the amplified risks that come with increasing complexity, including exposure to malicious inputs and generation of undesirable outputs. Discover practical strategies for combining tools from Arize, Databricks MLflow, and Mosaic AI to evaluate and enhance high-performing agents, drawing from extensive real-world experiences across numerous productionized use cases. Understand the essential relationship between agent development and evaluation techniques, where both must simultaneously improve to drive effective self-improvement. Gain insights into managing the growing sophistication of autonomous agents while maintaining their reliability and safety in production environments.