At Instagram, we serve billions of users worldwide, making our Android app one of the largest and most actively developed in the world. With thousands of code changes every day and weekly/biweekly releases, maintaining performance at scale is a constant challenge. Among these challenges, memory management stands out as critical—unchecked memory issues like leaks, oversized JVM objects, or native memory bloat can trigger OutOfMemory (OOM) crashes, Application Not Responding (ANR) errors, and other reliability problems that directly degrade user experience. Our data shows a clear correlation between memory inefficiencies and user engagement, making optimization a key lever for success. In this session, we’ll share how we’ve operationalized memory management at Instagram’s scale across core surfaces like Stories, Reels, Feed, and Messenger. We’ll dive into: Memory issue categorization and pipelines: How we classify and prioritize leaks, native memory growth, and JVM bloat using automated tooling. Triaging at scale: Strategies to identify root causes efficiently across a rapidly evolving codebase, including OOM crash analysis. Dynamic, memory-aware components: Techniques to make features adaptive to memory constraints (e.g., gracefully degrading resource-heavy features during low-memory conditions). Automated testing: How we integrate memory checks into CI/CD to catch regressions before they ship. Lessons from the trenches: Real-world examples of memory issues, including high-profile OOM incidents, and how we resolved them. Join us to learn how we balance rapid iteration with stability, ensuring Instagram remains performant even as we push the boundaries of what’s possible on Android.