Growing up in Oklahoma, Brandon D. Anderson didn't have any ambition to become a tech entrepreneur.
"I was a queer, black and homeless teen. I didn't even have the mental space to dream." He struggled, but at least he was with his "soul mate," another homeless teen whose name Anderson wouldn't share at the request of the man's family.
Anderson enlisted in the Army around 2003, keeping his love life a secret as required by army policy at the time. Then in 2007, he suddenly came out to his front-line supervisor.
He had to, he said, if he wanted to see his partner again. The man was hospitalized and lay dying after a traffic stop in Oklahoma City turned tragic.
"The police said they thought he had stolen a car. He had not," Anderson soberly recalls. "He was in the hospital for some time. They beat him first. They killed him."
The experience inspired Anderson to become a dedicated community organizer, helping other military veterans dismissed from the Army for their sexuality seek honorable discharges, and later organizing discussions around police brutality.
He's built a chatbot, Raheem.ai, that gives the public a way to grade police interactions as easily as they would rate and review a product on Amazon.
"Police departments crunch huge amounts of data today," he told CNBC, "but we still don't know how often law enforcement officers have hurt, killed, or for that matter saved and comforted people in the line of duty."
The lack of transparency contributes to a widening rift between police and the communities they promise to protect, he said. This is expecially true in communities of color. For instance, independent studies by the Guardian have found that police in the U.S. killed black people at twice the rate of white people in both 2016 and 2015.
People access Raheem.ai through Facebook Messenger on smartphones, or the company's own website. The chatbot asks users to answer simple questions about a police interaction, covering everything from basic details about where and when it took place, to more qualitative aspects, such as whether the officer made you feel "heard" or "disrespected." The bot asks users to select from a checklist of options but allows room for some details to be submitted through written answers.