Modern mobile apps must do more than just work. They must work smartly, even when offline. (Think of common applications like Twitter or Instagram.) This talk will walk through building an AI-enhanced, offline-first architecture on Android using custom vector stores. You’ll learn how to integrate on-device LLMs, enable smooth bi-directional sync with backend stores, and adopt scalable syncing patterns used in real-world apps. We’ll also cover sync conflict resolution, data serialization formats, and an efficient approach to user-local data modeling. All without sacrificing performance.