A U.K.-based startup is harnessing the power of visual artificial intelligence to help automate early-stage process development as well, they say, as biomanufacturing.
Reach Industries, which gave a Dragons’ Den talk at the 21st Annual bioProcessUK conference last year, has developed Lumi™ to include a similar technology to popular image-recognition apps, such as Brickit and PictureThis.
“The way data is captured in science hasn’t fundamentally changed since scientists first began scribing in notebooks,” explains Joe Mawby, business development manager at Reach. “Now we’re making waves with Lumi by capturing visual data with cameras and then detecting what’s happening with visual AI.”
AI trained on database of science images
Lumi works on similar principles to Brickit, a consumer app for suggesting Lego builds from piles of bricks, and to PictureThis, an app that can recognize tree health and species using a mobile phone image. Instead of trees and bricks, however, Lumi uses AI trained on a database of science-relevant images to recognize further images of the same type.
Reach Industries can train the AI, Mawby explains, to fulfill a variety of purposes. For example, Lumi might detect the color change of a fluid flowing through the viewing window. Alternatively, it could act as a second pair of eyes for a lab protocol where, currently, two technicians need to sign off quality control, he says.
“Fundamentally, anywhere there’s a form of human assessment involving visual data, Reach can support organizations,” Mawby says.
To aid the capture of visual data, Reach offers small “wifi enabled” cameras to their clients. However, Mawby explains, they can also use their own cameras with custom software and, later this year, the company also plans to offer a wearable version of their camera.
“How the visual data is collected doesn’t matter, provided it can feed into the platform for use by the visual AI,” he points out.
However, despite offering a wearable camera, Mawby explains the data output isn’t a heads-up display or augmented reality. Instead, it looks more like the software output from the popular Ring doorbell and security systems.
“You can select an area for constant output, and it gives constant feedback, but that’s sent to your Chromebook, iPad, or similar, rather than appearing in front of your eyes,” he explains.
Currently, Lumi is available for early-stage process development. However, Reach Industries hope to offer the system for use in GMP-compliant industry at a later date.
