In this hands-on workshop, you will learn to deploy Google's Gemma-3n models directly in web browsers using MediaPipe's LLM Inference API. We'll build privacy-first applications that run completely on-device with WebGPU acceleration, starting with text generation and progressing to Gemma-3n's multimodal capabilities that process text, images, and audio inputs simultaneously.

We will also explore advanced features like LoRA fine-tuning for custom models, all while ensuring user data never leaves the browser for complete privacy and instant responses.