Database Reference
In-Depth Information
This was my first experience coding in C—finding memory leaks—and I kind
of enjoyed it. I had a huge bank of Mac Pro towers, and I got to parallelize code
across all the towers. It was just a real blast. During this experience I started
thinking that there were some aspects to the “real world” that I could be
interested in. However, I still didn't really know where to go with it.
I ended up briefly at the NSA for a summer internship program called the
Director's Summer Program. They basically get a bunch of math nerds, put
them in a room, and throw food at them while they do math. I did that and
realized I didn't really want to be in government. Not because they don't
tackle important problems—they do, but because I encountered a lot of folks
who were just waiting to retire. For example, I worked with one person who
had a picture of a golf course up above their desk who would say, “This is me
in two years.” I was hoping for more excitement than this in my work life.
Gutierrez:
Having ruled out government and academic math, what did you
do upon graduation?
Foreman:
I went into consulting, in part, because I wanted to be able to
attack problems at a fast pace. So I joined Booz Allen and started doing a lot
of different analytics projects for the government. The team I worked with
mainly dealt with quick-hit analytics projects. For example, one of the projects
was around BRAC [Base Realignment and Closure] for the Army. Since the
Army can't shut down training when it moves bases, we modeled the move of
the Armor School, with all of their tanks and heavy vehicles, from Fort Knox
to Fort Benning with respect to training demands. Our mathematical models
allowed the military to train personnel to operate heavy artillery all while
shutting part of one base down and relocating it. Don't ever let someone tell
you that you can't change the tires on a moving car!
The great thing about doing analytics for the government is that there are
lots of opportunities to improve things. They've got a lot of very complex
problems that have very interesting constraints. For example, how do you pre-
dict tax return fraud when you're not allowed to use discriminatory features,
such as zip code, which are quite correlated with the fraud itself? There's a
challenge!
From government, I moved into consulting for large enterprises doing pricing
models, and some blending models for juice products over in China, which
was a blast. So at this point, I had had some large enterprise experiences, but I
wanted to get a sense of data science in the startup world. One of the things
that I wasn't given a chance to do a whole lot of in the enterprise world was
supervised artificial intelligence. I did a lot of forecasting, which is a type of
predictive modeling. You're using time-series data, along with a lot of optimi-
zation modeling when that's around pricing or supply chain, but not a whole
lot of true machine learning.
