High-performance cloud computing startup Paperspace Inc., which offers virtual desktop services with a focus on media-rich applications and 3-D graphics, is now taking aim at machine learning.

The company has just launched Gradient, a new cloud-based development platform designed for building, training and deploying machine learning models. The Gradient service runs on Paperspace’s graphics processing unit-accelerated desktop as a service and virtual desktop infrastructure platform in the cloud.

It’s aimed at students, data scientists and nontechnical professionals who don’t have access to the expensive computing infrastructure that’s necessary for machine learning. The idea is to provide a robust suite of tools that are easy to use, so those users can incorporate machine learning into new applications they build.

Paperspace reckons its new service can help to make the use of machine learning and other cognitive technologies much more widespread. One of the main problems holding back these technologies is the immaturity of the tools and frameworks necessary to enable them, the company said.

“Gradient makes high-performance cloud computing less expensive, more powerful and easier to use than competitors by making it readily available on the cloud, and for everyday users,” Daniel Kobran, co-founder of Paperspace, said in an interview with SiliconANGLE. “Gradient focuses on simplifying the tools and streamlining design and user experience, empowering anyone to better leverage the incredible gains of artificial intelligence and machine learning.”

The service allows users to launch a GPU-backed Jupyter notebook directly in their web browser, providing an interactive environment that combines simple code with media tools including visualizations, equations and rich text.

Within that environment, users can review the data they intend to use as the basis for their new machine learning models, then quickly start training that model using either a base algorithm from TensorFlow or one of Paperspace’s preconfigured algorithms. Once trained, the platform allows for the new models to be incorporated within software applications with just a few clicks.

“Keep in mind, users don’t have to build a model from scratch,” Kobran said. “Even the most sophisticated machine learning folks today are often building on top of a standard set of models. Paperspace has some templates available with existing models to get you started.”

Image: Paperspace/Facebook