Google AI Edge Open-Sources ML Drift GPU Compute Engine
Google AI Edge has open-sourced ML Drift under the Apache 2.0 license, providing a high-performance, cross-platform GPU compute engine for real-time on-device AI and machine learning inference. By abstracting low-level GPU graphics APIs, ML Drift simplifies developers' ability to run complex generative AI models and real-time computer vision applications locally across mobile, web, and edge hardware. ML Drift supports OpenGL ES, OpenCL, Metal, and WebGPU backends, serving as the primary GPU acceleration engine inside LiteRT (formerly TensorFlow Lite) while also being available as a standalone library for custom graphics and inference runtimes.
## BACKGROUND
Running machine learning models directly on edge devices reduces latency, improves user privacy, and eliminates reliance on cloud infrastructure. LiteRT, the successor to TensorFlow Lite, is Google's core framework for deploying on-device ML and generative AI models across mobile and embedded platforms.