Database Reference
In-Depth Information
on. Sometimes I say, “Oh, I don't have those tools,” so I have to find the tools.
Either I build them myself or we have to find another away. This means we
can go out and find one or put it into the pool of tasks for our engineers. One
of the engineers may then take it and build it, or come back and tell me that
somebody must have solved it before, so then we'll dive into the literature to
start looking for a paper.
In regards to helping our team learn things, we engage in a great deal of lec-
tures. This involves me giving lectures to our group so that everybody can
learn more about machine learning. Some of the days I'm giving presenta-
tions, whereas other days I'm in the audience learning from someone else.
Sometimes it's theoretical topics and other times it's somebody more on the
engineering side of things talking about something he did or a new technology
he brought in. Here's the thing—everybody's learning something new every
day. And these lectures are fun for everyone. People start thinking in the new
terms they are learning when they try to solve their problems and use the
different tools they care about. So that's good.
In regards to meeting people, I also meet with VCs, customers, people we are
looking to hire, and people interested in data, data science, and big data, and
so on. It depends on the day and it can be quite diverse. Just a few days ago I
met the Israeli British ambassador, as well as the Israeli president, at a dinner
I was invited to about big data. Earlier this year, I went to the conferences at
MIT and Strata. So part of my job is to meet people around this topic. Funnily
enough, my husband is also in data science, as he has a data science startup.
In the evenings, when we relax, we talk about our life like regular couples, as
well as about our data science problems. It's a pretty regular life with a lot of
action and I love it.
Gutierrez: What specific tools and techniques do you use at work?
Radinsky: Sounds funny to say, but I use the tools and techniques needed
to solve our problems. Our production code is in Java and Scala. Most of the
stuff I do is written in Scala. When it goes into production, we usually write it
in Java because it's more maintainable. The prototypes I build will be written
in Scala. For our stack, we work on Amazon completely. If somebody wants
to try something new, I just start a server and it finds those things and scales
those things out there. For our data storage, we use MySQL because we're
pretty small and we have enough relational data. That said, for some of our
biggest data sources, we use NoSQL databases, such as Couchbase.
The rest of tool set is all of the different algorithms that we have developed.
Many of our machine learning algorithms were written in the company as
some of the known algorithms were not appropriate for the types of prob-
lems we were handling. But we still play around with it a little bit. I even have
people writing in R. R is not very popular here because we write all of it in
Scala, so there are a lot of Scala toolkits for the things R or Weka are capable
of doing. We do a lot of NLP, so we use the Stanford Parser.
 
Search WWH ::




Custom Search