NEWS n 5 UK start-up uses AI to enhance production quality in real time A UK START-UP IS BUILDING industrial AI into factories, giving them the ability to understand and improve themselves in real time. London-based Matta was founded in 2022 as a spin-out from the University of Cambridge's Institute for Manufacturing, and has already deployed its technology in factories in the UK and Europe making products including waterproof coats, speaker cabinets and robot arms. Most of the existing installations are using vision systems to identify manufacturing flaws rapidly and easily, but the company has ambitious plans to expand its capabilities to fix production errors in real time and, eventually, to build its own automation platform. Matta commissioned its first installation earlier this year and is currently installing about two systems a month. It says it has a waiting list of 300 interested customers and hopes to be doing around ten installations per month by mid-2026. The company, which started 2025 with five employees, currently has 13 (drawn from MIT, Google X and Microsoft, among others), and expects to be employing up to 40 by the end of 2026. Rather than adopting a common academic approach to engineering, which starts with an idea and funding and then discovers the market does not need the product – Matta’s founders visited more than 50 UK factories, talked to engineers and operators to discover their issues, and then designed its system around them. Matta says it can typically set up its inspection system in less than half a day. Its cameras watch items on a production line for a few hours or days, while a proprietary AI model learns what “good” looks like. The system can work unsupervised (determining independently what is good or bad) – or be guided by human experts to capture the knowledge of a factory’s skilled workers. The AI learns the line like an apprentice, consolidates inspection, measurement and QC in one place, traces likely root causes, and helps to fix problems before they become costly. “Most factories already have the data they need – it’s just locked in people’s heads and on the shopfloor,” says Matta co-founder and CEO, Dr Doug Brion. “We’re using AI to turn that tacit know-how into something the whole factory can see, trust, and act on in real time.” Images from the cameras are captured and processed, and defects and anomalies are identified and logged in real time – often with such precision that it spots quality issues that had been missed previously. The system can check surface finishes and assemblies, as well as performing tasks such as measuring and counting. It is robust to changes in lighting, position, and other real-world variations. The system also creates a part-level record at every key step. This can be used to intercept defects before assembly. In the event of a warranty claim, a product's history can be traced to reveal, with images and data, whether the issue arose from a component, a process or misuse. www.drivesncontrols.com November/December 2025 The cameras can also capture barcodes or serial numbers, powering MES-like functions where objects can be tracked between stations. Bottlenecks and cycle times can be monitored in real time, all while qualifying every part in the factory after every process step. Matta claims that the capabilities that differentiate its approach include: n Fast, low-effort set-ups Its systems are quick to install and do not need perfect lighting or complex jigging. Matta takes care of all the hardware, integration, AI training and software – there is almost nothing for customers to do. n A single platform for vision and factory data Inspection and production data are linked so users see not only what has failed, but why. They get live visibility of a line, can trace parts, spot bottlenecks, and understand process issues – from one place. The technology brings together what would usually sit across MES, QC, ERP, and factory IT systems. n Flexible AI that handles real-world conditions Unlike traditional vision systems that depend on hard-coded thresholds and fixed set-ups, the AIbased system can deal with variations – such as moving parts, poor factory lighting, and unseen defect types. It can judge subtle issues – such as “is this scratch too bad?” – pick up multiple defects per camera, and scale to almost any part or process. If the AI system detects errors, the next step is correction, using a model that can predict the root cause behind the error, why it happened, and the corrective action needed to fix it. By partnering with machine OEMs, Matta is incorporating this capability into nextgeneration machines, with AI-driven closed-loop control systems that can fix themselves on-the-fly and learn how to use materials they have not seen before. Within 18 months, Matta hopes to be predicting potential problems, and then to deliver feedback that can guide product designs and material selection. Eventually, Matta hopes to bridge the gap between design and production. By bringing real factory data into the tools that engineers use to design products, Matta will help them build products that are scalable, reliable, and work right first time – even for entirely new and ambitious ideas. “If we can plug advanced AI into very ordinary factories, the impact on productivity and competitiveness could be enormous,” says Brion. “If it takes six months and an army of consultants to deploy, it is not a solution – it’s a science project. We aim for plug-and-play: a camera, a small box, and a model that learns the line like an apprentice. Matta’s AI-driven, vision-based quality control system can detect previously undiagnosed defects
RkJQdWJsaXNoZXIy MjQ0NzM=