Nervana Systems, a startup assembling data center hardware for an increasingly trendy branch of artificial intelligence known as deep learning, announced $3.3 million in fresh funding today.
Of course. Because producing hardware for specific purposes takes money — and becaused Nervana isn’t the only one looking at hardware for deep learning. That style of computing entails training systems called artificial neural networks on lots of information derived from audio, images, and other inputs and then presenting the systems with new information and receiving inferences about it in response.
“Deep learning is at the bleeding edge of academia, really,” Nervana chief executive and co-founder Naveen Rao told VentureBeat in an interview. “It’s a great place to hire people from.”
And talented engineers could be especially curious if they know major companies are involved. That seems to be the case at Nervana.
“We talked with a number of different potential partners and customers, which I can’t really disclose, but suffice it to say that it’s a lot of big names out there,” Rao said. “There is definitely a growing pain point for scalability of these solutions.”
At the same time, energy efficiency of hardware is critical, and indeed Rao is thinking about how much power gets burned on a per-computation basis.
Rao wasn’t ready to show a prototype yet. Those should be available for customers to try in the next six months or so, he said.
Draper Fisher Jurvetson led the round, and Steve Jurvetson will join Nervana’s board. Allen & Co., AME Cloud Ventures, and Fuel Capital also participated in the round.
Nervana announced its seed round just four months ago. To date it has taken on $3.9 million.
The San Diego startup employs 11 people and an intern now. Another six or so people will join over the next year, Rao said.
Even with the competition, Rao sounds certain he’s in the right business.
“We’re definitely getting a lot of feedback that it’s the right path to go down,” he said.
The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here