Deep learning involves a highly iterative process where data scientists build models and test them on GPU-powered systems until they get something they can work with. It can be expensive and time-consuming, often taking weeks to fashion the right model. New startup Determined AI wants to change that by making the process faster, cheaper and more efficient. It emerged from stealth today with $11 million in Series A funding.
The round was led by GV (formerly Google Ventures) with help from Amplify Partners, Haystack and SV Angel. The company also announced an earlier $2.6 million seed round from 2017, for a total $13.6 million raised to date.
Evan Sparks, co-founder and CEO at Determined AI, says that up until now, only the largest companies like Facebook, Google, Apple and Microsoft could set up the infrastructure and systems to produce sophisticated AI like self-driving cars and voice recognition technologies. “Our view is that a big reason why [these big companies] can do that is that they all have internal software infrastructure that enables their teams of machine learning engineers and data scientists to be effective and produce applications quickly,” Sparks told TechCrunch.
Determined’s idea is to create software to handle everything from managing cluster compute resources to automating workflows, thereby putting some of that big-company technology within reach of any organization. “What we exist to do is to build that software for everyone else,” he said. The target market is Fortune 500 and Global 2000 companies.
The company’s solution is based on research conducted over the last several years at AmpLab at the University of California, Berkeley (which is probably best known for developing Apache Spark). It used the knowledge generated in the lab to build sophisticated solutions that help make better use of a customer’s GPU resources.
“We are offering kind of a base layer that is scheduling and resource sharing for these highly expensive resources, and then on top of that we’ve layered some services around workflow automation.” Sparks said the team has generated state of the art results that are somewhere between five and 50 times faster than the results from tools that are available to most companies today.
For now, the startup is trying to help customers move away from generic kinds of solutions currently available to more customized approaches, using Determined AI tools to help speed up the AI production process. The money from today’s round should help fuel growth, add engineers and continue building the solution.