An Ottawa-based mould removal company is turning its machine learning solution over to the open source community in hopes of developing a robust platform to speed up the identification of dangerous moulds.
Mold Busters, founded 15 years ago by Ottawa entrepreneur Michael Golubev and his father, Andrey, is a mould identification and remediation company operating in Ontario and Quebec with franchises in a few Asia-Pacific countries. The firm recently unveiled its InstaLab platform, which looks to identify mould just by taking a picture of the fungal growth, potentially saving days on the current time-intensive processes.
“This was kind of our moonshot project to see how we could reinvent the industry,” says Michael.
The younger Golubev’s career took him out of the mould treatment business and deep into 3D printing for a number of years, even earning him an appearance on the TechCrunch stage a few years back. But when he returned to lead Mold Busters as its CEO three years ago, Golubev hadn’t lost his passion for disruption.
The problem with mould removal, he tells Techopia, is a common one in traditional, service-based industries: a lack of innovation. For the past three decades, identifying mould has been done the same way: taking test samples on-site, shipping them in cassettes to a lab and waiting a few days for a mycologist to tell you what you’re looking at.
“The process is very slow. We’re trying to see how we can speed it up so anyone can take a microscopic image and, right away, identify the mould type,” Golubev says.
If Mold Busters’ algorithms are properly trained, anyone in the industry might be able to take a photo of the mould they’re facing, cross-reference the sample with a database to determine the exact fungal family it belongs to and then set straight to work remediating or removing it.
While the promise of the InstaLab platform, currently in beta, might be revolutionary, the grunt work required to upload, identify and tag photos of thousands of varieties of mould is immense. Instead of doing all that labour in-house, Golubev says he took a lesson from his previous startup, 3Dponics, which solicited designs for 3D files to print parts for hydroponic gardens.
Mold Busters has opened up all of its data and algorithms to the open source community to feed the database with as much information on mould as possible. While that leaves Mold Busters without a firm claim on the resulting intellectual property, the InstaLab platform can quickly grow through others training the system.
Within the first few days of putting the word out about the platform, Mold Busters heard from research labs in the United States that were excited to share their work to bolster the project.
“So anybody can take this and work with it. And we just encourage that everybody contributes to the data,” Golubev says. “As the dataset grows, the system gets stronger.”
Going open source doesn’t mean Mold Busters won’t gain business from the platform. Golubev says the time and resources saved by not sending samples to research labs is invaluable, and will give the company a competitive edge over traditional competitors. Further down the line, the company could also develop a mobile app with the platform embedded to make the program more accessible to the industry.
For Mold Busters, which employs 12 people in Ottawa and 45 more around the world, the InstaLab platform is also a chance at history. To become a certified mycologist in North America, candidates have to take a test that involves correctly identifying some 80 per cent of mould samples pictured. If the InstaLab AI becomes robust enough to provide that level of reliable identification, it could conceivably pass that test, Golubev says.
“If the stars align, we can have the first mycologist-certified AI on the market.”