Background checks can be an expensive, time-consuming process. Now, a startup hopes to apply machine learning to provide a trust rating on people in less than a minute for under $1 per check.

Trooly Inc. today announced the official launch of its Instant Trust rating, which uses the branch of artificial intelligence that allows computers to learn somewhat akin to humans rather than being explicitly programmed. The two-year-old, 21-person startup, based in Los Altos, CA, collects data from public sources such as search engines, social media, national sex offender registries and crime records and runs them through machine learning algorithms to determine how trustworthy a person or business is for a particular job, loan or other activity.

To date, the company has mostly gotten revenue–several million dollars so far–from companies in the sharing economy. Trooly won’t name most of the customers yet, though one is UrbanSitter Inc., the babysitter startup. That focus stems from the reality that current background checks take too long, cost too much for some sharing economy applications and often don’t actually catch behaviors the companies want to avoid.

“Existing trust mechanisms in our society weren’t working for the New Economy use cases,” Trooly cofounder and Chief Executive Officer Savi Baveja. Indeed, the cofounders, who also include Chief Technology Officer Anish Das Sarma and Chief Science Officer Nilesh Dalvi, ran into just some of those situations personally, such as a rental and a ride-sharing experience that didn’t meet quality promises.

Pilot tests

The company also is in pilot tests with financial services firms to verify people’s identities and predict the value of potential customers. They’re also used in employment checks in industries where official background searches aren’t required but are useful, such as house cleaning.

Trooly’s Instant Trust ratings won’t replace conventional background checks in most cases, especially those where they’re mandated by law, such as ride-sharing services. Instead, they’re intended to serve as additional checks before and after the usual background screening.

They can be used initially to assess whether people are likely to pass a traditional check and if so, the company can focus more extensive background checks only on others deemed less likely to pass. They also can be used for continuous monitoring after the conventional check.

“We’re able to surface more areas of concern than you can get in a background check,” said Baveja, and those areas can be specific to certain jobs. For example, anti-social language detected on social media might be less important for someone applying for a ride-sharing job than for a babysitter.

Privacy concerns

Although Trooly uses publicly available records, the mining of all this data raises obvious privacy and regulatory concerns. The company avoids tapping into Facebook profile data, which the social network doesn’t allow without permission even for public profiles. It also uses machine learning algorithms to avoid associating people with a crime they didn’t commit but which appears on the same page as their name. And the algorithms can distinguish between an assault and a speeding ticket and between recent and older crimes.

The company’s coming out of stealth after starting to offer its services last August because Baveja believes “we’re now at the stage where the models are really robust.” Trooly also announced today a Series A round that brings the total amount raised by the company to $10 million, from Bain Capital Ventures and Milliways Ventures.

Besides competition from standard background checks, Trooly is up against other startups in the field such as Onfido and Checkr Inc.

Image from Trooly