Startups: Focus on DeepCube, FortressIQ

By AI Trends Staff

In a periodic profile of selected startups, we feature a look at DeepCube and FortressIQ.

DeepCube: Shrinking Deep Learning Models to Boost Performance

Startup DeepCube this week launched a software-based inference accelerator said to improve deep learning model performance on any existing hardware.

DeepCube was co-founded by Dr. Eli David, a professor at Bar-Ilan University, Tel Aviv, Israel and a leading expert in the field of computational intelligence, specializing in deep learning and neural networks. He is also the co-founder of Deep Instinct, which applied deep learning to cybersecurity.

Dr. Eli David, co-founder, DeepCube

DeepCube’s chairman and other co-founder is Yaron Eitan, a serial entrepreneur who has built and supported many technology companies in the past 30 years.

Asked by AI Trends in an interview to describe the DeepCube team, Dr. David said the 15 AI researchers and developers in the company “are the best students I have ever had,” with masters and doctorate degrees.

Asked to describe the business problem DeepCube is trying to solve, Dr. David said many Fortune 500 companies are getting good results in their research on AI proof of concept projects. “Their problem is deployment, not only on the edge but in the cloud. They are paying exuberant amounts of money to execute in the cloud,” which also presents latency, bandwidth and privacy issues, he said. These obstacles are holding AI back. “You don’t see wide deployment of AI, despite all the hype,” Dr. David said.

The DeepCube accelerator reduces the footprint required by the inference engine to do its work, making it “10 times faster with less memory, resulting in a 10x reduction in the cost of computing,” Dr. David said. Using patented technology, DeepCube is said to continuously restructure and “sparsify” – to make leaner – deep learning models during the training phase. The goal is to reduce the size of the model while maintaining or improving accuracy.

Neural networks and deep learning models are becoming “heavy,” said co-founder Eitan, requiring big infrastructure investments to run. “We can provide advanced computing for a fraction of the cost,” he said.

Yaron Eitan, co-founder, DeepCube

The company’s go-to-market strategies include working with developers of AI hardware accelerators, licensing directly to enterprises who want to employ in their own environment or in the cloud. “We don’t care,” where the deployment happens, Dr. David said. The third strategy is to work with vertical solution providers, such as makers of security cameras, drones and possibly, self-driving cars.

The company has raised $10 million in the last two and a half-years.

Learn more at DeepCube.

FortressIQ Discovers Business Processes Ripe for Automation

FortressIQ offers the Process...