Efficient Machine Learning, Automated

Model Compression

Deliver orders-of-magnitude model size and latency reduction with the same or even higher model accuracy. Supports major ML tasks, such as 2D/3D detection, face/pose recognition, GAN, 2D/3D segmentation, etc.


Train an efficient model with state-of-the-art AutoML techniques, push-the-button solution for hardware-aware model design.

Fast Inference Engine

Fast inference engines optimized for mobile devices, 3x faster than TensorFlow Lite Micro. Ready to use solutions on STM32, Raspberry Pi, and Jetson devices.