An Agent That Fixes Its Own Mistakes - Saloni Gandhi & Ziwei Guo, LinkedIn

Why watch

Saloni Gandhi and Ziwei Guo from LinkedIn demonstrate how AI agents continuously improve themselves through a closed-loop system. The talk covers GEPA (Generic Evolutionary Prompt Adaptation) which enables agents to evaluate their outputs, identify failure patterns, reflect on root causes, and propose targeted fixes. This is a deep-dive for engineers building production AI systems who want to implement self-correcting capabilities in their agent architectures.