Determined AI Inc., a new startup set on easing artificial intelligence development, today exited stealth mode with $11 million in fresh funding.

The round was led by Alphabet Inc.’s GV investment arm. A half-dozen other backers participated as well, including storied venture capital firm CRV, SV Angel and Amplify Partners.

Determined AI has built a software platform that aims to take some of the hassle out of setting up a deep learning development environment. One of the main tasks it promises to simplify is infrastructure management.

Enterprise AI projects tend to require significant amounts of hardware, particularly for the training phase of development in which algorithms’ accuracy is honed. Determined AI orchestrates hardware resources using a cluster manager that was created specifically with deep learning in mind.

According to the startup, it’s several times more efficient than the general-purpose tools enterprises normally use. The software is said to be capable of performing hyperparameter tuning, the process of optimizing the configuration settings that determine how an AI processes data, at up to 50 times the speed of other products.

“Many companies employ cluster schedulers like Kubernetes, Mesos, or YARN, which can be used to run deep learning workloads,” Determined AI founders Neil Conway, Evan Sparks and Ameet Talwalkar wrote on the startup’s blog. “However, traditional cluster schedulers and leading DL [deep learning] frameworks have been designed independently, which results in both poor performance and usability.”

Layered over the platform’s infrastructure management features is a set of development tools. Engineers can automatically generate neural networks from predefined specifications, experiment with different algorithm variations and create AI training workflows.

The platform also takes into account the collaborative nature of most deep learning projects. A built-in version control system enables AI teams to save training workflows for reuse, while labeling features makes it possible to organize experiments in a way that lets engineers search their colleagues’ projects.

Companies can deploy the platform both in the cloud and in data centers on-premises. According to Determined AI, early customers have been running it in production for more than a year now to support their deep learning projects.

“Our customers tell us that we have already saved them hundreds of engineering hours per person per year and hundreds of thousands of GPU-hours across their teams,” Determined AI’s founders wrote.

Photo: Unsplash