Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.


Latvia headquartered Enot, an AI startup backed by New Nordic Ventures, has announced the launch of a dedicated framework to optimize deep neural networks.

The AI-driven offering, as the company explains, comes with a Python API that can be integrated within various neural network training pipelines. It then automates the search for optimal network architecture, considering various hardware and software-centric parameters, including RAM, latency, model size constraints and operation type.

“Enot.ai’s framework takes a trained neural network as input, after which it selects a subnetwork that has the lowest latency and can ensure no accuracy degradation,” the company told VentureBeat. 

The whole process makes the neural network faster, smaller, and more energy-efficient, solving the challenges commonly faced by developers worldwide.

Event

MetaBeat 2022

MetaBeat will bring together thought leaders to give guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.

Register Here

Enot’s Impact

Enot claims that its solution has the potential to help AI developers and companies achieve up to 20 times neural network acceleration and up to 25 times network compression. The benefits even help cut down computing hardware costs by as much as 70%.

“Enot is at the forefront of next-level AI optimization, helping bring fast, real-time levels of AI advancement … Our journey has only just begun with examples such as the Weedbot laser weeding machine that gained 2.7 times acceleration, thanks to the Enot framework,” Sergey Aliamkin, CEO and founder of the company, said.

Overall, the company claims to have run pilot projects with over 20 companies, including major players such as PicsArt, LG, Huawei, Dscribe and Hive.aero.

In one case, it accelerated an image enhancement neural network by 13.3 times for a smartphone manufacturer. The optimization reduced the neural network depth from 16 to 11 and reduced the input resolution from 224 x 224 pixels to 96 x 96 pixels, without any loss of accuracy, the company said. It also had another project with the same company where the framework delivered 5.1 times acceleration for a photo denoising neural network, without any change in quality.

“Before meeting us, they already had several customers, including large international tech companies like LG, Huawei, Sony. That confirmed for us that Enot is solving a business-critical issue in the neural network space that cannot be solved internally nor are there any feasible solutions available in the market,” Dmitry Saikovsky, General Partner of New Nordic Ventures, told Venturebeat. 

Other players looking to solve the same problem are Deci.ai, OctoML, DeepCube, Deeplite, NeuralMagic, and DarwinAI.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.