Research to Reality with Google DeepMind — Paige Bailey, Google DeepMind

Why watch

Paige Bailey demonstrates Gemma 4, Google DeepMind's open model family, showcasing local inference running entirely in-browser and on mobile devices. The talk covers model sizes from 2B to 31B parameters, quantized checkpoints under 1GB, Android/iOS support via Google AI Edge Gallery, and the practical benefits of running AI locally without API calls or data leaving the device. Developers interested in on-device AI deployment will learn about choosing appropriate model sizes and hardware acceleration options.