From prototype to production: shipping AI features in your Android app

Getting Gemini working in your Android app takes an afternoon. Getting it ready for real users is where things get serious: exposed API keys, abusive clients burning through your quota, prompts frozen inside a released APK, and no visibility into what your AI feature is actually doing in the wild. This session tackles the production gap. Using Firebase AI Logic, you will call Gemini directly from Kotlin with no backend to build, then harden it for launch: block unauthorized clients with App Check, update prompts and model configuration server-side without shipping a new APK using Server Prompt Templates, and track usage, latency, and cost with AI Monitoring. Finally, we will look ahead to hybrid inference: how the same API prepares your app to run Gemini Nano on-device, so your AI features can work offline, respond faster, and cost less. Ship AI features you can trust in production, and be ready for what comes next.