Published in


OmniML Emerges from Stealth to Empower Edge AI Everywhere with Ease

OmniML’s technology focuses on hardware-aware model compression before training, which unlocks massive potential for performance and efficiency gain

We are excited today to introduce OmniML which is accelerating artificial intelligence (AI) on edge devices by making machine learning models smaller and faster, while also improving inference performance across any type of hardware. We have a vision where AI on the edge creates new possibilities to improve the quality of our lives at home and work — and everywhere in between.

OmniML is headquartered in San Jose, Calif. and is currently hiring to expand our growing team. We are looking for thought leaders, scientists, engineers, bold doers, and fun people to come join us at OmniML. Join a culture of collaboration where ideas are valued, different perspectives encouraged, and bold creativity greatly appreciated. Check out our open positions and more information at

Behind OmniML, there has been years of research, innovation, and proofs of concept — not to mention founders who are accomplished in researching and engineering new AI technologies and techniques. OmniML is led by co-founders Dr. Di Wu, the company’s CEO, Dr. Huizi Mao the company’s CTO, and Dr. Song Han, MIT EECS professor and serial entrepreneur whose groundbreaking research demonstrated neural architecture search techniques that can drastically reduce neural network size without losing accuracy.

Dr. Han’s work on efficient inference engines first exploited pruning and weight sparsity in deep learning accelerators to help lay the groundwork for OmniML. The OmniML team also includes world-leading researchers and industry experts from Stanford, MIT, UCLA, CMU, and other successful AI start-ups and Fortune 100 technology companies.

OmniML Secures $10M in Seed

OmniML has secured $10 million in seed funding to achieve its mission of making edge AI easily accessible to people everywhere. The seed round is led by GGV Capital with additional investment by Qualcomm Ventures, Foothill Ventures, and other prominent venture capital firms.

The time is truly here to revolutionize AI, and OmniML is leading the charge by ensuring AI can be accessed closer to where most people live, work, and play.

OmniML Technology

OmniML’s technology has helped many customers streamline their AI/ML model creation process and let the machine engineers focus on the core ML algorithms for their cameras and other edge devices while OmniML optimizes and enables advanced computer vision (CV) models to be easily deployed, maintained, and enhanced at the edge. Its software enables AI/ML algorithms to run on the individual devices with no need for additional local processing or battery power and eliminates the need to send data for analysis and inference to the cloud or data centers.

OmniML” Compress” the Model Before Training — bringing HW deployment constraints into model design and training

While there are newer cameras with specialized ML-enabled chips to accelerate ML training and AI inference, they come with higher energy consumption requirements, and are typically cost-prohibitive for most applications and use cases.

This is where OmniML enters the picture and empower your devices with advanced AI capabilities

One of the main use cases that OmniML has helped customers tackle is in Advanced Driver Assistance Systems (ADAS) space. OmniML is working with one public EV company, and one self-driving robotics company to ensure their AI/ML models are optimized and enhanced when deployed on self-contained vehicles.

3D Detection and sensor fusion with multiple cameras

None of this will be possible without more flexible, power-efficient, low-latency edge AI solutions that are embedded or easy-to-install, and which perform at levels that meet or exceed what can currently only be done in the cloud or data center.

OmniML’s AI/ML technology automates model co-design, training, and deployment targeting graphic process units (GPU), FPGAs, AI system-on-chip (SoCs), and even tiny microcontroller units (MCU) like those found in edge and mobile devices.

OmniML enable efficient AI — creating the best models on different platforms effortlessly

Because our technology works on any hardware, it not only alleviates the need for dedicated programming for each platform, but it also provides an easier upgrade path when devices need to take on more AI capabilities.

OmniML’s technology is based on industry-leading, edge deep learning TinyML technique dubbed MCUNetV2. MCUNetV2 has outperformed other leading ML methods running on microcontrollers with higher accuracy rates for detection and additional vision applications that were previously impossible.

OmniML’s core technology has won several awards, including:

As we continue to work with customers and partners to test and deploy AI on the edge, we invite you to check back in to learn more and evaluate where you can implement edge AI to create new insights, deliver new value, and gain new competitive advantage.

Come and be part of the edge AI revolution. We are hiring.

Learn more and follow us at:



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
OmniML Inc.

OmniML Inc.

OmniML is an enterprise artificial intelligence (AI) company that aims to effortlessly empower edge AI everywhere.