Large vision models give Quality Control a Big Brotherly spin

Ever wondered how your evening snack-time cookies turn out perfectly uniform in shape and texture? Chances are they were inspected by the cameras and AI systems built by this Chennai-based startup. Jidoka Technologies’ software tools and computer vision models help detect product defects in real-time across assembly lines. From biscuits to car gears to coconut oil bottles, its systems use multiple cameras and powerful vision models to flag anomalies and eject defective items in a flash as the products run through conveyors.

“Modern computer vision can transform product inspection as manufacturers aim for higher quality,” says Sekar Udayamurthy, co-founder and CEO. Typically factories engage workers for manual quality checks. But, they are subjective as individual workers cannot perform uniformly causing either higher wastage or higher defects. Also the process could be slow. The startup makes it consistent, faster and helps companies scale inspections. It can inspect 10,000 parts per minute with more than 99% accuracy, according to Udayamurthy.

It monitors manufacturing processes involving human workers in real time to prevent rework. Its tools can be used for various tasks including counting, sorting, label text verification, assessing inventory and suggesting improvements in the manufacturing process with analytical dashboards. Jidoka provides hardware tools along with its software and models.

Giving us a lowdown on this entrepreneurial journey, Udayamurthy informs us that he was working in a senior role in the US at Cognizant, building products for Fortune 500 clients while his co-founders, Dr Krishna Iyengar, Vinodh Venkatesan were both employed in an auto manufacturing facility in Chennai. They identified quality control as a key pain point in the manufacturing sector, which led to higher wastage and costly recalls. So they began building industry-specific AI models.

Starting with automobiles, Jidoka Technologies performs three million inspections a day across FMCG, pharmaceutical, electronics and logistics verticals for major companies like Britannia and Maruti Suzuki. “We are the only ones who give both product and process inspection on a single platform. Secondly, we are the only ones that can do it at these speeds. These are our primary differentiators,” he said.

The trio has built domain-specific vision models for each industry and holds two patents in anomaly detection and decision engine. The startup was incubated by NSRCEL of IIM Bangalore and has so far raised

19.5 crore. It raised 11 crore in Pre-Series A round from Arali Ventures. Jidoka aims to hit the $25 million revenue mark in five years time and bets on expanding its services to different verticals such as solar and semiconductors as well as geographic expansion in Australia, Europe and North America. They plan to build a platform for proactive monitoring for quality and throughput and building large vision models (LVM) to handle tasks in multiple domains. “The end-game is to create generic models that can work across customer bases,” explains Udayamurthy.



Ready to Make a Smarter Property Decision? Build Your Legacy with TOI Homes.