Just how far out on the edge can edge AI go? A throng of technology companies are set to find out as the roll out an array of hardware and software products designed to allow AI processing and analytics in a wide range of IoT devices and end points.
The list includes Stream Analyze, an edge AI software firm based in Uppsala, Sweden that just announced an expansion into the U.S. market, where it already boasts a partnership with Chandler, Arizona-based Microchip Technology.
Stream Analyze proposes to convert previously "stupid" machines into intelligent IoT assets with its SA Engine software platform for edge analytics that the company says allows AI models to be developed, deployed, and evolved on any device. If you are wondering about the list of devices that might include, let your imagination wander to not just connected vehicles and smart factory equipment, but also beyond that to items like lawn mowers and even chainsaws. If that sounds strange, consider that those machines, like many others, have engines and other components that are sources of device health data, and that can benefit from predictive maintenance.
Almost any machine nowadays can be fitted with sensors and connectivity to gather and transmit data off-device to be analyzed, but Stream Analyze told Fierce Electronics that some devices are just to small, and that AI capabilities are hard to deploy in a scalable and economically sustainable way, particularly when data needs to be sent to the cloud for AI analysis. Doing so can result in energy, computing, and storage costs that far exceed the cost of the hardware itself. Installing Stream Analyze’s SA Engine and relevant AI models on edge devices will allow companies to reduce those costs, while letting them brainstorm new business models that can leverage the presence of edge AI, the company said.
The platform can be used on devices and microcontrollers already deployed in the field, provided they have the computing resources. “Such devices include Arm Cortex M4 and M0,” Stream Analyze said via email. “Even though others have successfully deployed neural networks etc on these, Stream Analyze is unique in that our software SA Engine enables interactive queries and model lifecycle management. This gives our users the opportunity to see, understand, and model their data more effectively. The interactive model evolution ability enabled by SA Engine helps our users accelerate their edge AI projects to market. Applications include acoustic and vibration analysis, as well as motor control and powertrain anomaly detection.”
SA Engine is not the only edge AI software platform around for IoT devices and applications. For example, AWS Greengrass, which Stream Analyze clearly has honed in on as the competition, also offers analytics capabilities via cloud and edge. But, according to Stream Analyze, SA Engine requires only 17kB of memory, making it 5,600x smaller than AWS Greengrass, while averaging 2x faster inference than TensorFlow Lite. Maintaining compact code and faster inference speeds will allow to leverage the value of analytics, the company said.
As for Stream Analyze’s full approach, SA Engine is the main component installed on each edge device, while SA Studio is the company’s front-end development environment. There is also SA Staging, a testing environment for model scaling simulations, and SA Federated Services, which facilitates integration with customers' existing infrastructure.
The nine-year-old Swedish firm already has several large international customers in sectors like automotive and manufacturing, including Volvo Group, Iveco, Toyota Material Handling, Autoliv, Husqvarna, and ifm, and is now looking to the U.S. to lengthen its list of partners and customers.
"Expanding into the U.S. market is a milestone for Stream Analyze, as it opens doors to a hotbed of innovation and technology," said Jan Nilsson, co-founder and CEO of Stream Analyze. "Our edge AI solutions are ideally positioned for sectors at the forefront of digital transformation, like the automotive, manufacturing, and chip industries, spurred by advancements like Tesla's disruption and Industry 4.0. This move isn't just about growing our footprint; it's about catalyzing a wave of innovation, making edge AI accessible, and driving significant business advancements through data insights.”