Paul Duan wants to change the way we think about data

Posted by on August 08, 2014.

FWD.us sat down with Paul Duan, the 22-year-old data scientist and founder of Bayes Impact. Paul gave us the inside scoop on his plans to revolutionize the public sector and change the way we think about data.

Paul-cover-FINAL.png

FWD.us: What is Bayes Impact?

Paul: We are a non-profit that deploys teams of data scientists to work with civic and non-profit organizations to create data-driven solutions to tough social problems.

FWD.us: So, can you explain what “data-driven solutions” means?

Paul: There are all of these big problems that nobody is really tackling. I used to be a data scientist at Eventbrite and I was working on fraud detection. If you have a better algorithm that can reduce fraud by 5-10% -- that’s good, and you can save your company millions of dollars. But what if you could use the same kind of techniques to achieve similar improvements when the metric isn’t a company’s bottom line, but human lives? A good example is emergency responders. If you could use an algorithm to predict where an emergency will be, you can get emergency responders to the emergency faster. Right now, if you had a heart attack and called a Lyft, the Lyft will probably get to you faster than the ambulance.

Obviously these are not directly comparable situations, but the bottom line is that Lyft can predict demand and use this information to help optimize their dispatch system. Similarly, predicting where emergencies are more likely to happen can help us reduce response times for emergency responders. This is a clear example of where the tech industry has developed a solution that can be applied to the public sector.

FWD.us: Can you give us more examples of what you do?

Paul: Sure. Consider the example of organ transplants: In the U.S. there are currently more than 100,000 people waiting for an organ, but people get added to the list about twice as fast as organs become available and the result is that about 20 people die on average every day because of the shortage of organs. But if we could better predict who are the most critical patients and which transplants are most likely to be successful, then we could allocate the few organs we do have in a fairer and more efficient way. 

FWD.us: You’re from France, originally. How did you get to the U.S.?

Paul: I came to study abroad at U.C. Berkeley on a J-1 visa [an exchange-student visa], but then I won a scholarship to be a visiting scholar instead, which meant I had to switch to an F-1 visa [a student visa]. This seemed like an insignificant change at the time, but when I finished my year at Berkeley and wanted to go work in the tech industry, I learned the hard way that the F-1 visa actually has different restrictions than the J-1 in terms of work eligibility. That was my first brush with immigration. Next, I had a company offer me a job and to sponsor an H-1B visa [a foreign worker visa], but unfortunately I got the offer right after the H-1B visa cap was reached. So, I was prepared to stay and work in the United States, and suddenly I had to leave everything and go back to France.

Paul_infographic_FINAL.png

I was later able to come back to the U.S. on a one-year J-1 visa, which was sponsored by my company, and hoped to be able to switch to an H-1B. But H-1B visas work on a lottery system -- and I lost the lottery. The company I was working for suggested I try to get an O-1 visa [a visa for individuals with extraordinary achievements or ability], which is difficult to qualify for. To bolster my application, I started entering some data science competitions and won a couple awards from Amazon and the Economist. I was written up in a few news publications and finally had enough to qualify for the O-1 and get back to the US.

I quickly got stuck again because O-1 visas are tied to your employer. So, when I wanted to leave my company to start Bayes Impact, I had to sacrifice my visa. I’m now in the process of getting a new one, but that’s a whole different story. 

FWD.us: How did being in Silicon Valley affect your idea for Bayes Impact?

Paul: What’s good about Silicon Valley is that an investor can give $1 million in seed money to a company, and that company can go on to create a $1 billion company. You can create value, because entrepreneurs build products and services that can scale. In the non-profit sector, you can compare that to multiplying impact. Not many people in the non-profit sector are really looking to use data to increase their impact; regrettably, non-profits are often more about redistributing money than they are about creating value through scalable solutions. I think in fifteen years, you will see most non-profits run very differently, and I want Bayes Impact to be on the forefront of that shift. 

Paul-Quote-Final.png

FWD.us: What excites you about the work you do?

Paul: We haven’t quite seen the scrappy and entrepreneurial approach to non-profits yet. We’re lucky, because as the only non-profit currently at Y Combinator [a tech incubator in Silicon Valley], we are encouraged to explore an entrepreneurial side of non-profit work. What’s exciting about data is that you can find the commonalities between all of the problems you have with the public sector, and create general solutions to them. You can have an algorithm running right now, in production, that optimizes the way organ transplants are prioritized in the US by predicting the likelihood of rejection. It’s not just a lofty idea -- it’s something you can actually ship in the form of a tangible software product. And when you’ve done that once, nothing prevents you from implementing a similar solution somewhere else. To make a nonprofit like Bayes Impact work, we rely on the fact that we can get top talent -- people who would be really difficult to get in the for-profit space. We can sell them something totally different. We can tell them: Do you want to optimize the way people click ads for the rest of your life? Or do you want to change the world?

 

Join the Movement Fight for Immigration Reform Today

Your information is secure and will not be shared