“Breaking down the barrier of machine learning”
An interview with Jan Jongboom - Co-founder of Edge Impulse
Watching your raspberry pi recognize objects after just a short time spent creating and teaching the model is jaw dropping. A machine learning (ML) platform like this is something that both the commercial and hobbyist communities have been looking for since artificial intelligence and machine learning became common topics of discussion. Edge Impulse is the company behind this software and have proven themselves as one of the next great startups in tech.
While working for Arm Ltd. on IoT devices (such as smart TV’s, wearables, etc..) in 2017, Jan Jongboom made an observation that would change the course of his work. He noticed that instead of helping people to understand the world, IoT devices were being used for mundane tasks such as checking temperatures or product availability. He realized these devices were essentially performing the same task over and over again with slight variations. To do complex tasks—such as determining if an animal is an elephant, for example—required multiple sensors and a surprising amount of code. He knew there was a better and more efficient way using machine learning (ML).
Jongboom’s realization sparked an idea for an ML project and ultimately led to the development of a new company, Edge Impulse. Jongboom began working on this project with his future co-founder Zach Shelby. They were able to get an easy-to-use working product—in partnership with Google—that focused on perfecting the user experience (UX) and eliminated much of the difficulty in converting sensor data to usable data. The project was created as completely open source and built off of existing technology. In this way, they were able to connect the dots instead of reinventing the wheel.
The next step for Edge Impulse was funding. They first held an angel investing round which included the co-founder of Hackster.io, Adam Benzion. Benzion subsequently stepped in as active Chief Experience Officer (CXO) of Edge Impulse. With Benzion stepping in as CXO and having completed another successful round of funding, it became clear that the company was headed in the right direction.
As Edge Impulse began to promote viable products, their commercial customers were able to drive revenue. However, in the commercial space—especially with big companies—customers already have an idea of what they want. It’s easy to accidentally become more of a consultant firm by only following the road maps of customers. Edge Impulse worked carefully to tailor products to fit customer needs while still conserving their own vision.
Along with commercial customers, Edge Impulse has an expanding community of DIY innovators and hobbyists. This can be largely attributed to the sheer number of tutorials and resources the company has provided. Edge Impulse has now grown to over 40 employees. However, since Jongboom is the most knowledgeable about the product he created, he still does many of the tutorials himself.
With a couple tutorials under your belt, building a machine learning model can take just minutes through the Edge Impulse platform. Technology like this has been trapped behind a barrier of education preventing many people from attempting to learn or use it. However, with Edge Impulse the technology behind ML and AI is being exposed to a world who has waited for it for quite some time.
Edge Impulse’s newest product, FOMO (Faster Objects, More Objects), is an ML algorithm that creates accurate models using almost 30x less processing power than other popular programs. This makes FOMO perfect for anyone working with devices that can’t handle the strain of a bulky ML model.