Google AI Edge stacks like LiteRT have powered massive Google and third-party edge applications globally, from leading apps like YouTube and Google Meet, to platforms like Android, Pixel and Chrome.
In this talk, we will share how Google AI Edge is pioneering on-device AI innovation to deliver transformative experiences across our products. Let’s explore our product suite to integrate on-device AI across mobile, web, and more to keep data private, latency low, and enable offline capabilities. We'll dive into the Google AI Edge stack, showing you when to drop in ready-made features using MediaPipe Tasks, powerful language models like Gemma, how to use LiteRT to deploy and accelerate your own custom models across platforms, and demonstrate our on-device AI native apps like AI Edge Gallery.
Key Takeaways:
Understand edge AI use cases, and related tiny / small models like Gemma4 E2B.
Learn Edge AI stacks like MediaPipe and LiteRT / LiteRT-LM, and use them to build on-device apps with hardware accelerations (CPU/GPU/NPU) through simplified APIs.
Build and deploy on-device agents with Gemma4, and use related tools to streamline your development journey .
Speaker
Shuangfeng Li
Senior Staff Software Engineer @Google AI Edge, 20 Years of Experience in Google's Core Engineering & Research Groups
Shuangfeng is a senior staff software engineer in Google AI Edge team, leading developer experiences for edge AI. His team has piloted on-device AI innovation, from open-source edge AI frameworks like LiteRT (previously known as TensorFlow Lite) and LiteRT-LM, Google’s OSS Gemma on-device models, to edge AI apps like AI Edge Gallery.
Shuangfeng has 20 years of experience in Google's core engineering & research groups, like Core ML, Google Brain, Search, Core Data and Google Maps etc. Particularly, previously in Google Brain, he has accelerated the TensorFlow community in APAC, and led TensorFlow Lite DevX team to improve on-device ML usability, infra, performance and hardware ecosystem.
