A startup called OmniML stepped out of the shadows today, armed with $10 million in seed funding to pursue its mission of accelerating artificial intelligence on low-powered devices at the network edge.
The seed round was led by GGV Capital, with participation from Qualcomm Ventures, Foothill Ventures, Matrix Partners, Tectonic, GSR and IMO.
OmniML is pursuing the goal of making AI more accessible and widespread, which it’s doing by developing smaller and faster machine learning models that can run on devices with less power. It says it’s solving the “fundamental mismatch” between AI applications and edge hardware.
The problem is that some devices, such as smart cameras and drones, lack the computing power to support large AI models and process information at the edge. So edge devices cannot achieve the same kind of autonomy that is possible in data centers and cloud environments, where apps have access to far more processing power.
OmniML said its compression software allows developers to create smaller and more scalable machine learning models that enable edge devices to perform AI inference much faster than was previously possible. As a result, it claims, it can perform multiple machine learning tasks up to ten-times faster on a variety of edge devices.
OmniML co-founder and Chief Executive Di Wu said many AI models today have grown so big that edge devices just aren’t equipped to handle their computational power.
“But that doesn’t have to be the case,” he said. “Our ML model compression addresses the gap between AI applications and edge devices, increasing the devices’ potential and allowing for hardware agnostic AI that is faster, more accurate, cost effective and easy to implement for anyone, in any setting.”
Using its software, developers will no longer need to optimize machine learnings models manually for specific devices and processors, meaning they’ll be able to deploy faster and with higher performance, the company said.
The company said its AI compression software is already being used by customers in areas such as video surveillance, creating more powerful smart cameras that have superior real-time situational awareness. Its software can also be put to use in self-driving vehicles, the company claims, as well as the precision manufacturing industry, where it helps to improve quality control detection models.
Although OmniML has revealed little about how its technology actually works, it claims to have won over a lot of fans. Amazon Web Services Inc.’s AutoML Library and Meta Platforms Inc.’s PyTorch deep learning framework have both integrated its neural architecture search algorithm, the company said. Its technology has also received a number of awards, ranking first in the Sixth AI Driving Olympics at ICRA 2021 and winning a number of other contests.
OmniML said it will use the funding from today’s round to expand its machine learning team and expedite software development.