This Penn startup uses ‘ethical AI’ to avoid racial profiling in shoplifting monitoring

Percepta, a year-old security startup out of University of Pennsylvania, is making moves this summer — namely in that four of its five cofounders are graduating from undergrad next week and are joining the Pennovation Accelerator in June to take next steps with business and development of their AI-based retail security product.

The tech uses what the cofounders call “ethical” artificial intelligence to anonymize shoppers’ race, gender and age for loss-prevention teams.

Instead of using regular video surveillance, the AI tracks shoppers’ body movements — scanning for employees, looking up for cameras or putting products in their clothing. Then, an image of a person the tech found highly probable of having shoplifted is sent via a mobile app to the store’s loss prevention team, who ultimately make the decision to intervene or not.

“It’s a social good, but it’s also a competitive advantage,” CEO Philippe Sawaya told Technical.ly. “We’ve seen that the public perception of AI is that it can be biased, and the public just isn’t going to accept that.”

Take a look at the product’s demo here:

Cofounders Sawaya, Neil Gramopadhye, Jonathan Mak, Alexander Lee and Rahul Shekhar began working on the concept in various classes during their time in undergrad. Sawaya, Mak, Lee and Shekhar will all graduate next week from Penn Engineering and the College of Arts & Sciences, while Gramopadhye will enter his senior year in the fall.

Sawaya and Gramopadhye, the company’s head of bizdev, said the team is currently working remotely, but is excited to access Pennovation’s space once the accelerator begins. Mak, Lee and Shekhar hold the titles of COO, CFO and CTO, respectively.

In the last year of developing the tech and the business model, the team has leaned on Penn professors Vanessa Chan and Jainbo Shi as well as Walmart VP of Energy and Facilities Mark Vanderhelm.

“Working on this startup has been a pretty amazing experience,” Gramopadhye said. “I’d never really expected this to happen, but it’s been one of the most interesting and dynamic experiences I could have had.”

The startup has raised $77,ooo so far, including funding from the Penn Wharton Startup Challenge at the beginning of the month, which Sawaya said will last the team a while.

The CEO said believes that this type of technology can be adapted for high-sensitivity uses like in medicine, or with tracking social distancing with the coronavirus.

Sawaya also said the team will work this summer toward putting more final touches on the technology of the product and starting marketing for it. They’ll be piloting the product first in smaller spaces, like the Penn book store, before they start to think about big-box stores like Target or Walmart.

“It’s part of our large vision that we want to be the new standard in retail,” Gramopadhye said. “When you see this store uses Percepta, you know you won’t be racially profiled.”