ProgrammableWeb first covered Ersatz earlier this year while the platform remained in private beta. Since then, Ersatz has continued to develop its neural network platform. Today, Ersatz launches an industry first GPU platform for deep learning with cloud technology. Prior to release, ProgrammableWeb caught up with Ersatz CEO, Dave Sullivan
Tech giants (e.g. Facebook, Google, Microsoft, IBM, etc.) continue to invest in deep learning projects and corresponding top talent. Until the Ersatz release, deep learning neural networks were only available to those with deep pockets. However, Ersatz aims to change that as Sullivan explained:
‘Off the shelf’ Deep Learning software does not currently exist, complicated by the fact that you really need special hardware (GPUs, Graphics Processing Units, specialized number crunchers) for Deep Learning to be practical. That has all changed with Ersatz.
Ersatz comes as a SaaS hosted version or as a Deep Learning Appliance for on-premise use. Both versions provide the needed GPU number crunching power behind the scenes along with a variety of Deep Neural Network algorithms that can be used to accomplish a variety of tasks. It gives engineering teams everything they need to start leveraging Deep Learning in their organization.
Today’s announcement furthers Ersatz’s aim to bring deep learning to the mainstream. Sullivan explained:
We are aiming to provide easy to use, scalable tools that enable companies to do Deep Learning without hiring a Deep Learning team. These tools are exposed to other applications via an API and can be used to accomplish a wide variety of tasks, e.g. medical imaging or algorithmic trading.
Predictive analytics, big data, deep learning, and other forward thinking technologies are yet to fully understand their entire potential audience. With that in mind, we asked Sullivan who the target audience with the Ersatz platform is. He answered:
Data scientists and engineers finding themselves dealing with machine learning problems more and more (and needing good tools to do so).
We are taking deep learning to the enterprise. We think that deep learning is going to move beyond the labs of data scientists to businesses that want to understand their data. We are providing a tool to use deep learning to analyze the data. Medical imaging and finance are two verticals that we are going after initially.
Sullivan found the beta trial massively successful. 2200 companies participated and Sullivan was amazed at the applications that were developed. Of particular interest, algorithmic trading achieved notably successful results. Now that the beta has ended, Sullivan explained Ersatz approach to the broader market:
We have the cloud version we are primarily promoting as a “first step” for using Ersatz via word of mouth. We are also selling the appliances, for those we’re trying to sell direct via b2b sales.
According to Sullivan, revenue will end up being the ultimate judge of success. However, Sullivan explained that revenue won’t come without a quality product and a solid relationship with the developer community. In closing, Sullivan left us with his vision for Ersatz going forward:
While Deep Learning is currently our primary focus, we aim to become ‘the intelligence layer of choice’ for developers. The opportunity in generalized machine learning is massive and Deep Learning is what’s currently propelling that forward.