Cartesiam helps builders deliver AI to microcontrollers – theinformationsuperhighway

Cartesiam, a startup that aims to bring machine learning to edge devices with microcontrollers, has launched a new tool for developers looking for an easier way to create services for these devices. The new NanoEdge AI Studio is the first IDE specifically designed for machine learning and inferencing on Arm Cortex-M microcontrollers that already power billions of devices.

As Cartesiam GM Marc Dupaquier, who co-founded the company in 2016, told me that the company worked very closely with Arm, as both have a vested interest in having developers develop new features for these devices. While the first wave of the IoT was all about sending data to the cloud, that has now changed and most companies now want to limit the amount of data they send and do a lot more on the device itself. And that's pretty much one of Cartesiam's basic theses. "It's just absurd to send all this data – which, incidentally, also makes the device accessible from a security perspective," he said. "What if we could do it a lot closer to the device itself?"

The company initially relied on Intel's short-lived Curie SoC platform. In view of Intel, it obviously didn't work out so well Support for Curie ended in 2017. Since then, Cartesiam has focused on the Cortex-M platform, which has changed for the better given its ubiquity. However, since these are microcontrollers with low power consumption, it should be noted that this is not about facial recognition or understanding natural language. Rather, using machine learning on these devices is about making objects a little more intelligent and, particularly in an industrial application, detecting anomalies or finding out when it's time to do preventive maintenance.

Cartesiam is already working with many large companies that build Cortex-M based devices. However, the NanoEdge Studio makes this development work much easier. "Developing a smart object has to be easy, fast, and affordable – and today it isn't, so we're trying to change it," said Dupaquier. However, the company does not try to present its product to data scientists, he emphasized. “Our goal is not data scientists. We're actually not smart enough for that. But we're incredibly smart for the embedded designer. We will solve 99% of your problems. "He argues that Cartesiam has reduced the time to market by a factor of 20 to 50" because you can get your solution up and running in days, not several years. "

A useful feature of NanoEdge Studio is that it automatically tries to find the best algorithm for a specific combination of sensors and use cases. The libraries it generates are extremely small and use between 4 KB and 16 KB RAM.

NanoEdge Studio for Windows and Linux is now generally available. Prices start at € 690 / month for a single user or € 2,490 / month for teams.