On-device LLMs, unlike remotely accessed models, unlock private-by-default, low-latency AI experiences that work anywhere, offline; ideal for mobile. In this talk, Michał will show how to run LLMs directly inside React Native apps using Vercel's AI SDK that provides a robust abstraction layer to simplify building AI applications. Same interface to run local and remote models. These capabilities are possible thanks to an open-source set of libraries we created. He'll dive deep into the provider architecture and demonstrate how we integrated it with the MLC LLM Engine and Apple’s Foundation Models available on their latest mobile devices.