With generative AI becoming more accessible, developers are exploring ways to embed it meaningfully into real-world applications. In this session, we'll dive into how we integrated Google's Gemini GenAI into a Kotlin Multiplatform (KMP) storytelling app to support both online and offline story generation—delivering powerful AI experiences across Android and iOS from a shared codebase.

This talk covers the practical architecture behind the integration, how MediaPipe’s LLM inference support was brought into Android, how a custom CocoaPod was used to run GenAI locally on iOS, and how we handled preference-based switching between local and remote generation. We’ll also discuss UI and UX considerations for generative experiences, and how to structure your shared and platform-specific layers for maintainability and testability.

Whether you’re building with KMP, exploring GenAI, or just curious about the future of offline AI on mobile, this session will give you actionable insights, real-world code examples, and architectural guidance.

Takeaways:

- A working pattern to integrate GenAI (Gemini) into KMP apps
- Using MediaPipe’s Gemini LLM on Android, and a custom CocoaPod on iOS
- Architecting AI features cleanly across shared and platform layers
- Managing API/local model switching and user preferences
- UI/UX patterns for GenAI-driven features on mobile

Audience:
Mobile developers working with Kotlin Multiplatform who are interested in integrating generative AI capabilities. Prior KMP experience helps, but the session is accessible to anyone with a background in mobile or AI integration.

This talk explores how to integrate Gemini GenAI into a real-world Kotlin Multiplatform app to enable both online and offline promt generation. Learn how to wire GenAI into Android with MediaPipe, run it on iOS via CocoaPods, and design a clean, maintainable shared architecture.