Qeexo adds AutoML to STMicro MLC sensors to speed tinyML, IIoT development

Machine learning developer Qeexo and semiconductor STMicroelectronics have teamed up to allow STMicro’s machine learning core sensors to leverage Qeexo’s AutoML automated machine learning platform that accelerates the development of tinyML models for edge devices.

Qeexo CEO Sang Won Lee told Fierce Electronics via email that as a result of the collaboration, users of STMicro’s machine learning core (MLC) sensors will not have to write their own code for their applications, or make adjustments in code to account to account for varying conditions and situations in complex application environments. That addresses a common barrier to rapid and widespread deployment of tinyML models at the edge, tinyML being an ML technique optimized for edge developments, including industrial IoT applications.

“Scalability is a major issue that we’ve seen since our early days of developing machine learning (around 2015),” Sang said. “Often, solutions are so specialized for one use case or environment and device that you can’t really migrate it elsewhere for reuse. For example, even if we are monitoring the same model of elevator, one that’s installed in Building A has different environmental variables from one that’s installed in Building B, and the important ML features [and] parameters can be very different – so an ML model developed for the elevator in Building A cannot be used in Building B. Also, seasonal changes in temperature and humidity can also require different models or improvement on models. All of this is extremely cumbersome to calibrate (often having to rewrite) by hand. Qeexo AutoML is able to automate machine learning development to greatly increase the productivity of engineers.”

Qeexo is targeting AutoML at IIoT applications such as anomaly detection, condition monitoring and predictive maintenance, as well as logistics and other IoT use cases. Sang said AutoML currently supports ST’s MLC sensors and Arm Cortex M0 to Arm Cortex M4-powered devices, which the CEO described as non-specialized, relatively small and affordable chips that are widely used in any industry. 

In teaming with STMicro, Qeexo’s AutoML advantages can provide added value on top of the low power consumption and other advantages that the STMicro MLCs provide. The two companies previously worked together in 2018 on finger-based sensing technologies using AI and ML.

The support for STMicro’s MLCs is part of the latest release of Qeexo AutoML,which typically gets a new release every two to three months, Sang said. Also in the latest release is super for a new type of machine learning classification - Multi-class Anomaly.

Qeexo’s technology is sensor and processor-agnostic, so the company is aiming to work with other companies. In addition to STMicro, Qeexo has partnered with Renesas, and also supports Arduino devices on its platform. 

RELATED: Qeexo and STMicro promote FingerSense tech