Swim.ai offers machine learning at the edge

I know I’ve been writing nonstop about the pendulum shifting back to computing on the device (so-called edge computing) as opposed to in the cloud, and this week’s startup profile is yet another example of why this shift is happening. Swim.ai launched this week with a data analytics software that can derive meaning from data on the fly.

Swim.ai CTO Simon Crosby is someone I’ve followed for almost a decade, as he worked at security startup Bromium and before that, at hypervisor tools company XenSource. Crosby has a penchant for brightly colored shirts and an easy way of explaining tough technological challenges. When we talked about Swim.ai and why he founded the company, he explained that the amount of data that sensors will pick up in the world is impossible to send to the cloud.

We have to process it at the edge.

We also need ways to make sense of data that can work across many different types of devices and do so quickly, without a lot of data wrangling. Swim.ai’s new software, SWIM EDX, purports to do just that by learning what data is coming from a device and recognizing patterns in it. The software takes up 2MB, which is too much to run a low-power sense, but reasonable for a larger edge computing device.

In Crosby’s example, a traffic light camera might produce gigabytes of data each hour, all of which a city that has hundreds of cameras could never afford to ship to the cloud. Depending on the use case, the Swim.ai software could make sense of patterns it sees and then predict what will happen next.

From there, that data becomes a series of insights the city could provide for route planning to everyone from first responders to Uber. Or the company providing the traffic cameras could offer route planning as a service. Swim.ai’s goal is to sell the software to companies that are building edge devices, public utilities and other service providers trying to offer insights on their own platforms of devices and to enterprises in various verticals that might want to use EDX on their own gear. There are a lot of options here.

One important aspect of Swim.ai’s software is that devices running it communicate with each other without having to send data to the cloud. These peer-to-peer networks are essential for edge computing and intelligence because they offer compute power and resiliency. Swim.ai has been around since 2015, but the time has come for the type of edge architecture it’s offering.