How SwitchOn automates quality check on the factory floor

(L to R) Aniruddha Banerjee and Avra Banerjee, co-founders, SwitchOn. Image: Selvaprakash Lakshmanan for Forbes India

Aniruddha Banerjee and Avra Banerjee are cousins-turned-co-founders with their first software venture, Abee Research Labs, that has found some early success within two years of launching their first product.
 
As engineers, they collectively bring more than a decade of experience to their startup in areas, including computer vision, Internet of Things and machine learning, and stints in hi-tech engineering companies such as Nvidia—the leader in computer processors for artificial intelligence (AI)—LSI, Schneider Electric and Samsung Electronics.
 
They discovered that quality inspection processes at some of the world's largest manufacturers are ripe for innovative interventions, as much of those remain manual. To that end, they're building an AI software-based platform at their five-year-old venture under the brand name SwitchOn, integrating camera modules, sensors and so on to help automate sorting out of defective products on the factory floor.
 
“If you have, for example, a product that has a defect, you essentially have a very bad impression about the brand itself, and even more so for premium products,” says Aniruddha. “Imagine you bought, let's say, a Tresemme shampoo… you definitely would not want their artwork to have a crack in the middle, right? Or some kind of a wrapping issue.”
 
In working with most of their customers, the founders saw a defect rate of at least 3-4 percent, he says. Therefore, “every minute they are shipping out about five to 10 defective products per line, so just imagine how big the problem is”.

The Banerjees recently raised $4.2 million in Series A funding, taking their total to $5.2 million. They have focussed on FMCG and automotive clients to start with, and some 20 early customers include ITC, Unilever and SKF. While they are ploughing back everything into building the business, and haven’t turned a profit yet, they are growing fast.
 
This year, on the back of some large orders, sales, which rose 300 percent last year, could grow as much as 10-fold, he says.

Also listen: SwitchOn's founders on how they're helping ITC, Unilever and SKF automate quality checks on the go
 
Aniruddha’s Nvidia experience includes being part of a team that was working on applications in autonomous vehicles. This brought plenty of exposure to the manufacturing industry, where they saw the problems around quality inspections.

Avra worked at the R&D department at Schneider Electric, located on the outskirts of Bengaluru. He was also among those who literally took a moonshot, working on a lunar lander and a rover at the early Indian private space tech venture Team Indus. He worked on areas, including avionics and electronics, there.
 
Teaming up with his cousin for SwitchOn, “we moved into a 2 BHK and got started, basically”, he recalls.
 
Now, with the fresh funding, and the backing of investors, including Axilor and Pi Ventures, they are ready to invest more in product development to improve customer experience, and add richer features and functions around quality management, reporting, root-cause analysis and so on.

“Also, we are looking to now slowly expand into the US,” says Aniruddha, who has previously worked in the US, at Nvidia. The US will be SwitchOn’s main market, apart from India. “The amount of value we can add there is a lot higher, because the quality spend and the number of defects are also a lot higher in the developed economies,” he adds.
 
With manual processes today, on assembly lines that see hundreds of units of a product every minute—like the shampoo bottle, or a small auto component—manual checks are both inefficient and ineffective, the entrepreneurs say.

Also read: Right to Repair: When can we stop shopping and start repairing?
 
SwitchOn involves a combination of off-the-shelf hardware, including cameras and other related electronic gear, and the proprietary software that the Banerjees have developed.
 
This connects with other hardware at the assembly line and on the shop floor to be able to physically separate out defective products at a much higher rate, they say. They’ve been able to eliminate a much larger proportion of defective products, faster, saving time, effort and money for customers.
 
“After installing our system, our customers are able to get 100 percent quality guarantee by the end of the line,” Aniruddha claims. SwitchOn also claims 30-50 percent reduction in overall quality inspection costs once its platform is up and running.
 
While computer vision has existed for 20 years, everything that's happened in the last four to five years with the emergence of what are called GPUs (graphics processing units) and the cost of silicon coming down (not counting the shortage caused by Covid) have made it feasible to build a cost-effective AI-based quality check solution for manufacturing customers, Avra says.
 
SwitchOn also takes into account that factories are often in remote locations with sketchy internet. “All our solutions are edge-first solutions,” he says. “The actual software and deep learning are running on the edge (meaning mostly on the device itself), and that's what allows us to do very high speed applications.”

The entrepreneurs have assembled a team of some 25 engineers and expect to go to 40 soon. They believe the additional features and incrementally making their product even more affordable will help in growing faster.