Performance issues in production mobile applications are often complex, time-consuming to diagnose, and require deep expertise across multiple domains, including mobile development, observability, infrastructure, and performance engineering. One of the key challenges is that each operating system has its own behavior, execution model, and diagnostic traces, requiring specialists to interpret the underlying signals and accurately identify the root causes of performance regressions. This talk presents a novel approach that leverages multi-agent AI systems to autonomously detect, analyze, and identify the root causes of performance regressions in real-world mobile applications. Beyond identifying regressions, the system is capable of correlating performance issues with recent code, configuration, and experiment changes that may have contributed to the degradation. Drawing from hands-on experience optimizing GoTo’s navigation platform, the session will showcase real-world examples where memory allocation was reduced from approximately 10 MB to 61 KB a 99.4% reduction in memory overhead while also achieving a 99.7% reduction in unnecessary UI recompositions. The talk explores how specialized AI agents can collaborate to replace traditional manual debugging workflows with automated, scalable, and repeatable performance investigations. Attendees will learn how multi-agent systems can continuously monitor production environments, reason across telemetry signals, correlate changes, and accelerate root-cause analysis, ultimately enabling engineering teams to resolve performance issues faster and with greater confidence.