Caitlin Smallwood - Data Scientists at Work - page 22

market, there's also a legitimate business need of moving fast if we really want
to keep our awesome business thriving.
Gutierrez: How would you describe your work to a data scientist?
Smallwood: I would say we're a team that does all kinds of statistical model-
ing. We really focus and output three things as a team. We work on predictive
models using all of the techniques that people in this field would be familiar
with—regression techniques, clustering techniques, matrix factorization, sup-
port vector machines, et cetera, both supervised and unsupervised techniques.
A second thing is algorithms, which I would say are obviously closely related
to models, except that they're embedded in some sort of ongoing process, like
our product. And then the third thing is experimentation and all the scientific
methodology behind that, which we leverage, as well as all the analytics that
go with each experiment that we run.
Gutierrez: How would you describe your work to a non-data scientist?
Smallwood: I would say that we collect the data on all of our customers
about how and what they're watching. Then we hunt for patterns in the data,
which we can then leverage to recommend things to them that they might
want to watch and essentially improve the service for everyone. So we lever-
age the information across the whole population to really make each of our
customers happier.
Gutierrez: How did you come to your current position at Netflix?
Smallwood: My first job was as a programmer with a consulting company. I
wasn't even really doing anything intense algorithmically. Through this work, I
happened upon an optimization problem and that's what got me interested in
operations research or OR. The interest was high enough that I went back to
graduate school to study OR. After grad school, I went to work for a smaller
company focused entirely on building OR kinds of models for different indus-
tries and companies. The company was later purchased by PwC. And that was
great, but it was right around the time when the Internet boom started, and so
that changed everything in terms of how opportunities just exploded.
I feel like I just kind of lucked into this career that happened to coincide with
the Internet. Suddenly it was like, hey, there's all this data that wasn't available
before, whole new opportunities of types of things you could build models
for, and whole new problems that need to be solved. Things you used to have
to “model” did not need models anymore, because there was so much data.
All you needed to do was figure out the median of some particular dimension
value. So that changed the whole world of opportunities.
As this boom happened, I became really interested in working on a product.
This led me into the space of creating analytic products, like personalization
engines or other components of products. That was a pivotal point in my
transition from operations research into algorithms and data. Really being able