Microsoft Corp. unveiled an array of new cloud services at its Build developer conference today aimed at enabling organizations to apply artificial intelligence more effectively.

The lineup is headlined by an experimental offering called the Brainwave Project. It provides access to specialized chips within Microsoft’s cloud data centers that have been configured for the sole purpose of running deep learning models and, consequently, can vastly outperform regular silicon.

The chips in question are field-programmable gate arrays from Intel Corp.’s Startix series (pictured). FPGAs can be customized all the way down to their circuitry with optimizations designed to speed up specific tasks. In the case of Project Brainwave, Microsoft equipped its processors with a homegrown AI engine and components designed to ease model deployment.

As a result, the service lets developers harness the chips without having to use the highly specialized software tools normally required to do so. They can instead employ the popular TensorFlow AI development framework or Microsoft’s rivaling Cognitive Toolkit, both of which are open source.

Microsoft plans to add support for more deep learning engines over time, as well as bring Project Brainwave to Azure Stack systems. These are appliances sold by partners that enable companies to create what is essentially a copy of its cloud platform in their data centers.

The move to add support for Azure Stack is part of a push by Microsoft to extend its AI services beyond the public cloud. Another key element of the plan is Azure IoT Edge platform, a set of analytics tools that can be deployed on connected devices such as industrial systems to analyze important data locally. Microsoft announced today that the offering will provide access to its Cognitive Services lineup, starting with the Custom Vision offering.

As the name implies, Custom Vision enables developers to create customized AI models for processing specific types of visual content. Microsoft this morning unveiled an upgraded iteration of the service that can be used to recognize individual objects within images.

The update is rolling out alongside smaller tweaks to a few of the company’s other computer vision services, plus an entirely new, experimental offering called Project Ink Analysis. Microsoft said that tool can automatically extract text and shapes from handwritten documents. It should come handy for organizations looking to reduce the number of people necessary to process physical records.

In conjunction with these updates, Microsoft is also enhancing the part of its AI portfolio focused on processing speech. The company has folded its previously disparate speech recognition, transcription, text-to-speech and translation services into a single application programming interface aimed at easing developers’ work.

Image: Microsoft