Latest News

Harvard Venture Story: Zoba uses machine learning and unconventional data to calculate risk and confront fear

1/24/2018

By Ralph Ranalli
Harvard Innovation Labs Content Editor

As a Marine officer working diplomatic security in Iraq, Dan Brennan knew all about the fear of the unknown. Too often, when trying to protect US diplomats in Baghdad from terrorists or other threats, he found that there just wasn’t enough reliable information to help security teams make intelligent choices about which routes to travel or what hot spots to avoid.

“I was frustrated because there didn’t seem to be any ability to make predictions about future events, or areas that were at high risk for future events,” he says. “And I thought that was a bit ridiculous with machine learning and all the technology that there was out there.”

That frustration marked the beginnings of Zoba, the Venture Incubation Program (VIP) team Brennan leads with his co-founder and younger brother, Joseph Brennan.

Zoba uses unconventional data and machine learning technology to create predictive models that can work in international and emerging markets, where conventional data is either unavailable or nonexistent.

The company recently received seed funding from a pair of investors (a formal announcement is coming soon) that will enable the Zoba to increase its staff to 8 employees and to roll out a demo version of its product within the next few weeks.

Dan and Joseph were close growing up in the Texas hill country north of San Antonio, but they had only a long-distance relationship after Dan left for college at Notre Dame. Joseph, who is four years younger, moved east to attend Harvard College, just as Dan started a 5-year stint in the military that took him to Iraq and Afghanistan. Joseph moved to Thailand and then China after graduation and was working on a master’s degree in economics at Peking University in Beijing when Dan called him about his idea for Zoba.

During their time abroad, the brothers learned that the difference between a risky area and a perfectly safe one could be a matter of a few blocks.

“There’s a lot of hype on the media, there are kind of blanket statements put out about the risk of different places,” Joseph says. “Being travellers ourselves and spending a lot of time out of the country, we thought a lot of that was very distant from the ground truth.”

Says Dan: “I think what we’re really trying to do is illuminate the real ground truth of risk around the world and make it so granular and so street level that you can make decisions off of it.”

Zoba does that by combining available, and sometimes unconventional, data with machine learning. One problem with predicting risk internationally is that data either doesn’t exist or isn’t publicly accessible. So instead, the company uses data from cities with good reputations for transparency and reliability.

“Then we take … events and match them with thousands of different variables in the environment, including population density, what businesses are the vicinity, what the street lighting is like, what the weather was like, what day of the week it was, and what time it was,” Dan says. “We take all these variables and we run them through machine learning algorithms to find out which ones had the most effect.”

Then they use that information to build models that they can apply to cities that don’t have data. In some cases, for example, population density may play a role in the likelihood of crime or other violent incidents. In a city where reliable population data isn’t available, studying cell-phone towers and where they are clustered can provide clues about density instead.

As the company has grown, the brothers have expanded their vision, and are working to apply their methodology not just to crime and terrorism overseas, which was their original focus, but also to other types of risks and ones that may be closer to home. The same technology that can assess risk for a terrorist event, they say, can also potentially tell local officials where opioid overdoses are most likely to occur, allowing first responders to better prepare.

That’s significant progress since they first started the company back home in Texas. Zoba really took off, they say, after Dan was accepted to the masters in public policy program at the John F. Kennedy School of Government and found the Harvard i-lab.

“The i-lab has been huge for us,” Joseph says. “When I think about Zoba, I tend to divide it between the pre-i-lab times and the post-i-lab times, they were much darker times.”

“They were dark,” Dan laughs. The brothers particular credit their relationship with their mentor, Christopher Geiger, who is part of the senior management team at Boston-based Modern Analytics, a company that specializes in predictive analytics for corporate clients.

“He’s really taken us under his wing and really mentored us through every stage of our business development … the business side, how to develop our markets, how to pitch people,” Dan says. “And then on top of that he’s really just made connections for us that we never would have made on our own.”

They also say they appreciate the support of other entrepreneurs in the Venture Incubation Program.

“When you’re working alone, you know, on Monday you’re saving the world and on Tuesday it’s all destroyed and you might as well just go home,” Dan says. “And I think being around other people who are going through the same trials and tribulations really evens out the experience.”

But the perhaps the best relationship that has emerged from their venture has been their own.

“We’ve really formed a friendship that maybe didn’t exist as deeply as it did before we started Zoba,” Dan says. “And it’s been really nice, our mom’s super stoked about it.”

“We joke that we never worry about company disputes because, if it ever gets, bad our mother will decide,” Joseph says.

Share This