Bitfusion has announced the beta availability of Bitfusion Flex, aimed at helping AI software developers address the complexity of clustering, sharing and scaling deep learning infrastructure, to speed the learning curve for building AI applications.
In business since January 2015, Bitfusion recently announced a $5 million Series A funding round led by Vanedge Capital.
Bitfusion Flex, sitting atop the firm’s core virtualization engine, aims to optimize deep learning workflows through the AI application lifecycle, including development, training and inference. It can be deployed into any data center or cloud and aims to enhance the use of GPUs, field-programmable gate arrays (FPGAs) and other AI compute resources across all major deep learning frameworks.
The core virtualization technology provides a transparent software layer that combines multiple systems into a single, elastic compute cluster that supports the sharing and scaling of compute resources. It is designed to work out of the box with pre-existing applications.
“The promise of AI and deep learning is immense, but in practice, leveraging GPUs, FPGAs and other compute architectures for performance presents huge challenges for AI developers,” said Subbu Rama, co-founder and CEO of Bitfusion. By focusing on the unique needs around AI development and deployment, we are enabling companies to leverage these efficient architectures with less effort. This accelerates time to business value for AI initiatives.”
Rama said in many of his engagements, his team finds software developers have little experience building AI applications. “It is early stage; it’s a very new field. The hardware people may know how to program on the GPUs, but the software development people, probably not. The Googles of the world know what they are doing but 90% of companies are just starting. The level of expertise across the board is average. We want everyone to be able to use the platform.”
Moe Kermani, managing partner at Vanedge Capital, stated, “The data center is moving to a heterogeneous computing model where CPUs are augmented by specialized co-processors such as GPUs, FPGAs, and more. Machine learning is the major driver of this shift and Bitfusion is well-positioned to simplify how applications take advantage of this environment.”
Kal Freund, senior analyst for machine learning at Moor Insights & Strategy, stated, “Bitfusion sees an opportunity to provide tools to manage the machine learning development process while squeezing high utility out of the racks of GPUs needed to do the job. Bitfusion is the first to my knowledge to address this growing need.”
Release 1.0 is three to four months away, Rama said. Customers can pay as a percentage of their infrastructure costs, such as AWS services, or pay for a license. So far, he said the reception of prospects has been positive. “We’re solving an important problem. Some companies provide apps to do deep learning and some companies provide hardware, but no one is doing what we are trying to do. It’s a new problem. But AI is the new dot.com. Every company will be adding AI, to be able to provide huge value to their customers.”
– By John P. Desmond, AI Trends editor
For more information, go to Bitfusion.io.