.webp)
In a significant move signaling its entry into the U.S. market, Stream Analyze, a leader in edge AI solutions, announced a partnership with Microchip and other American companies to bring its innovative technology stateside — such as putting AI in chainsaws.
Founded in Uppsala, Sweden in 2015, and buoyed by decades of academic research, Stream Analyze is not just crossing geographical boundaries: it’s setting out to redefine the capabilities of hardware and equipment typically used with minimal or no software — yes, that includes chainsaws, but also lawnmowers, mining rigs, forklifts and other commonplace tools and devices.
“What we’re trying to achieve here is to make sure that our customers are doing what we call edge analytics, or edge AI,” said Daniel Spahr, Stream Analyze’s chief operations officer, whose LinkedIn bio neatly summarizes his mission as “Making stupid things smart,” in a video interview with VentureBeat conducted earlier this week.
Who needs an AI chainsaw?
What are the benefits of doing this? After all, who really wants or needs an AI-equipped chainsaw?
Well, according to Stream Analyze, the benefits are actually pretty easy-to-understand and significant: imagine a fleet of loggers dispatched to a clear a forest.
Wouldn’t it be nice for their manager or boss to know which chainsaws were running low on gas, how they were performing, if there was wear and tear and if one or several were likely to fail? That way, arrangements could be made to have one ready to switch out, reducing downtime.
It’s a big pitch for edge AI devices in general, including infrastructure sensors offered by rival Eloque, a company spun off of Xerox when I worked for the company a few years ago: detecting likely problems before they happen, improving efficiency, saving time and cost, and keeping operations running smoothly.
Another potential rival might be Sima.AI, which offers a no-code platform for edge AI and machine learning (ML) deployment, but is geared for potentially even more computational heavy devices such as military and reconnaissance drones. (That company just raised $70 million, showing the entire edge AI space to be one of potential interest to investors and VCs.)
The benefits of Stream’s end-to-end edge AI/ML platform
But Stream Analyze believes its tech is superior to others because it is an “end-to-end platform for machine learning op[erations]” according to Spahr, one that identifies the correct data to record and upload from the field back to cloud servers, without capturing needless data that would increase the cost, computational and energetic requirements.
“The cost for offloading data and storing it and using the processing centrally has become expensive,” Spahr explained to VentureBeat. “So you want to push things out locally to have that hybrid solution.”
There’s also the fact that in many edge AI applications, wireless connectivity itself may not even be an option — think about areas with poor mobile service coverage such as the bottom of a mine or out at sea.
“You might have connections sometimes and sometimes you have not,” Jan Nilsson, Stream Analyze’s co-founder and CEO told VentureBeat in the same video interview. “And in the meantime, the product is still being used. So the wear and tear is continuous. So you need to do some kind of analysis directly on the machine, on the device. We are fully independent of communication infrastructure.”
The suite of products offered by Stream Analyze includes the SA Engine, SA Studio, SA Staging, and SA Federated Services, which together provide a comprehensive platform for deploying AI models at the edge.
“We provide certain templates and models out of the box to the customer, but usually they build their own models,” using the SA Studio, said Nilsson.
The AI is typically deployed on specific hardware spec’d to the industry or company in question, often microcontrollers, which Stream Analyze designs for by consulting reference designs and working with partners in the semiconductor industry.
This system allows for real-time data processing and is designed to be user-friendly so not only data scientists but also analysts and engineers can effectively manage and deploy AI solutions.
According to Spahr, other rival technologies “often require an embedded programmer or firmware updates or something similar, which is either risky or slow or both.”
Rapid deployment and fine-tuning
In addition Stream Analyze claims its technology enables for rapid development, adjustment and maintenance of AI models directly on devices.
“It’s really easy to push out new models instantaneously onto a device, and this means that your time to market is sped up substantially compared to other technologies out there,” said Spahr.
With a footprint requiring as little as 17kB of memory—significantly smaller and faster than competitors like AWS Greengrass or TensorFlow Lite—Stream Analyze’s solutions are uniquely suited to a broad range of applications.
And Stream Analyze is content to allow its customers to tailor its technology to their uses without it ever even knowing about the end result — ensuring privacy and confidentiality.
“No customer or the other are alike,” Nilsson said. “They have different use cases, and they prioritize different things, so they use different models. There is no ‘one size fits all’ here. And in many cases, we don’t even know what kind of use cases they are addressing. They implement and deploy the platform, and then they they don’t share the models with us, so we don’t actually know what kind of analytics they are running.”
As Stream Analyze steps into the U.S. market, is expects to not only increase its business but also drive significant advancements in how businesses use data and AI to make real-time, informed decisions that were impossible before.
Venture Beat's original article →
