Moving to Palantir from Mathematics: Alumni Spotlight on Justin Bush

I recently finished my PhD in math at Rutgers University where I worked on topological approaches to dynamical systems. My interest in the field of data science was sparked when I first encountered topology being used to analyze real-world data. Although those kinds of applications were not the specific focus of my research, it became obvious to me with time that the kinds of problems data scientists are working on are exactly the kinds of problems I would like to work on.

Well, I got a job for one. Already by the second and third week of the Data Incubator there were companies contacting me that may not have noticed my resume so readily otherwise. I also got a tremendous exposure to the variety of data science jobs out there, something that would not have happened had I taken a job directly out of grad school.

Finally, I got to meet and became friends with the other Fellows, as well as everyone involved with the Data Incubator. It was encouraging to be surrounded by so many like-minded people, and it will be exciting to stay in touch with them as they go off to do interesting things in different industries.

I would start coming up with ideas for your project early and practice playing around with data in Python, especially if you haven’t before. Ideally you will find a compelling story that you are trying to tell, and let that story drive the data analysis and presentation that you do. In the end your project may or may not matter for getting a job, but I found it to be a great way to build confidence that I can do interesting work, and in relatively little time.

I think mathematicians get the benefit of the doubt as far as being technically competent. And in general, I think that presumption is pretty fair—there will be a lot of unfamiliar math and programming concepts when you start out, but having a mathematical background makes it possible to learn what you need quickly. The challenge is to convince people that you can apply that technical ability to real problems.

My project was based on New York City taxi data recently made available by Chris Whong. There were lots of directions I considered going, but ultimately I chose to focus on two things.

First, I wanted a measure of how long any given cab ride in the city at any given hour during the week could be expected to take. But not just an average—I wanted to estimate the distribution of times to get a sense of how long it would take, for example, in slow traffic. For trips where it’s important to not be late, knowing the worst case is more useful than just knowing what is typical.

Second, I wanted a visualization of how difficult it is to hail a cab at different times of day. This depended on inferring the number of available cabs in an area at each time of day, something not explicitly present in the data.

The natural way to convey this information was to build a website, which ended up being at least as much work as the analysis! It’s very satisfying, though, to build something so quickly and see that it actually works for the most part.