- Turing Labs, a small Toronto startup, is doing research and development for multi-national corporations that have closed their testing labs during COVID-19.
- The startup harvests a company's historical data and uses machine learning to simulate experiments on potential new products, with the goal of finding new formulas for the likes of soap, laundry detergent, or salad dressing.
- Tests that used to take a year and a half – and a room full of sweaty test subjects – can now be turned around in a little more than a month.
- Turing Labs is announcing a small seed round of venture capital funding this week and investor Eric Ries, author of "The Lean Startup," says that the company is a perfect example of the economical startup he wrote about.
- The CEO of Turing says that the startup is already turning a profit.
- Visit Business Insider's homepage for more stories.
A small startup in Toronto called Turing Labs is helping the biggest consumer packaged goods companies in the world do research and development during quarantine, replacing their sprawling laboratories with machine learning programs accessed on laptops.
Big laundry detergent companies normally test new products in vast labs that have rooms stuffed with sweaty test subjects dirtying clothes which will be shoved into their 250 washing machines. The companies experiment with fragrances and chemicals until they find just the right formulas. That's how consumers get detergent pods with the fresh scent of lavender, and the bright colors that make actors smile with approval on TV ads.
Those labs are empty now, and the test subjects are sweating it out at home as COVID-19 shuts down R&D facilities for many large firms. Yet consumers still do laundry, eat salad, and take showers (perhaps a bit less often, for better or worse). How do multi-national corporations test the detergent, salad dressing, and soap?
That's where Manmit Shrimali, the CEO of Turing Labs, and his entire staff of two employees come in. Shrimali and company harvest a firm's historical data from chemical and marketing tests to simulate experiments for hypothetical new products, applying machine learning to test a detergent in virtual loads of laundry that would ordinarily be done in person.
What if the lavender scent were replaced with sandalwood? How would the ingredients react chemically, based on how the chemicals interacted in past experiments?
What about a more peppery taste in that salad dressing – would the chemicals in the formula work and would consumers in the Midwest wince? Turing's system could quickly show that the chemicals would be fine, according to tests in which peppery flavor was added to another salad dressing, but that the Midwest appetite preferred a milder salad dressing formula two years ago. Shrimali says his platform can almost perfectly predict chemical interactions, but human preferences are more difficult, and companies will still need focus groups and other human testing.
These experiments normally take up to a year and a half, from sweaty subjects to new product formula. In Turing's software, the virtual sweat and chemistry takes just six weeks.
Shrimali listed off a handful of gigantic customers, but said that he couldn't name them publicly: "You would not believe the amount of legal paperwork that's involved with deodorant soap formulas."
But he says the startup has is overrun with demand and pausing new customers until August. To help it keep up, it just raised a scant $1.75 million in seed funding. Shrimali says Turing is already profitable and has job offers out to two potential candidates. The round, announced this week, was led by Moment Ventures, with participation from Y-Combinator and Eric Ries, author of "The Lean Startup," a Silicon Valley classic.
"'The Lean Startup' is all about using experiments and rapid iteration to find out what customers really want and need," Ries says. "That is what Turing is doing by using AI to make the product development cycle shorter and allowing companies to get higher quality products into the hands of consumers."
Ammar Hanafi, who led the round for Moment Ventures, said Turing Labs' technology is "proving critical for its customers to quickly adapt and keep innovating during this time."
Shrimali and cofounder Ajith Govind created the company after touring a major R&D center in the UK as consultants in the spring of 2019. There they saw the 250 washing machines — their destiny gleamed in the sweat of the test subjects.
"In the airport on the way back to Toronto we mocked up a prototype and emailed it to the company," Shrimali said. "'If we can do better than your lab in predicting how the chemicals interact, will you give us your business?' They said if we could do so with 90% accuracy, they would buy our software."
Instead of selling a one-off software, Shrimali and Govind started building their platform. Their software and testing begin at $50,000 – and they don't expect it to get any less popular when the COVID-19 restrictions lift. He guesses that the old labs will probably scale down considerably as R&D becomes more virtual.
"All the data and machine learning can be put into action along with the physical lab," Shrimali says. "This was probably coming for a long time, but like a lot of digital transformation, COVID speeded it up."