When Bitfusion released at Disrupt NY 2015, its target was on helping developers pace up their programs by supplying them pre-compiled libraries that created far better use of GPUs, FPGAs and other co-processing systems. That was two a long time back. Today, the best current market for these systems is in training deep finding out types, a thing that was scarcely on the radar when the firm released. Unsurprisingly, even though, that is particularly what Bitfusion is concentrating on now.
As the firm announced right now, it has lifted a $5 million Series A round led by Vanedge Money, with participation from new investor Sierra Ventures and present traders Information Collective, Resonant VC and Geekdom. The firm plans to invest this dollars into strengthening its exploration and development initiatives and to target on Bitfusion Flex, its new framework-agnostic platform for building and handling AI tasks.
Now in beta, Bitfusion Flex essentially aims to give developers a solitary platform for handling the lifestyle cycle of an AI application. Developers get a solitary dashboard that normally takes them from development to teaching, tests and — finally — deployment. Under the hood, Flex takes advantage of containers to make scaling and transferring experiments and types in between local equipment and the cloud simple, but it also supports deployments on bare metallic, also.
It’s crucial to note that Flex’s target isn’t necessarily on generating the modelling simpler. Whilst it does present an app retail store-like encounter for setting up your framework of choice (no issue no matter whether that is TensorFlow, Torch, Caffe or similar applications), its strength is in handling the infrastructure you need to have to develop and operate these programs. For the reason that of this, it neither cares about the framework, nor where by you want to deploy the application.
The provider provides each a web-based mostly interface to deal with this procedure as properly as a command-line interface that, for instance, lets you connect remote GPUs to your local notebook throughout the development section.
“A ton of individuals who commence deep finding out tasks simply cannot acquire them further than the prototype section,” Bitfusion CEO and co-founder Subbu Rama advised me. “Everybody desires to do deep finding out just about everywhere, but the World-wide 2000 — they really don’t have plenty of individuals.” So with Flex, Bitfusion wants to abstract the laborous function of handling infrastructure absent so that the facts experts that organizations do finally deal with to use can target on their programs.
Hunting ahead, Bitfusion plans to extend Flex and provide it out of beta in the following few months. The Austin-based mostly firm also plans to extend its Silicon Valley existence (even though Rama noted that most of the R&D function will however transpire in Austin).