Database Reference
In-Depth Information
that they know how to generate the process, and then talk through the entire
process of data science. I have a lot of questions around the tools and tech-
niques they would use and why they would use them. I want to hear the depth
of knowledge here, as well what works better, when it works better, and why
it works better. Afterwards, I usually tend to choose an algorithm they say
they know very well and ask them to dive in and explain the math.
Another thing that I am deeply interested in is people knowing data struc-
tures, like basic computer science. So I have started asking questions about
very basic data structure problems and then trying to move on to questions
around other data structures they don't know to see how they deal with it.
The last problem I give them is usually an out-of-the-box kind of question that
is impossible to solve. I want to see how creative they are in their approach.
This is important because I think a data scientist has to be creative in how
they solve things. I just want to hear how they solve something new and what
questions they ask.
Gutierrez:
What advice do you have for companies that are looking to hire
data scientists that don't have a data science team yet?
Radinsky:
The person you hire has to understand the business. I wouldn't
hire somebody who's a junior in the beginning. It has to be somebody who
not only knows the toolkits and techniques, but somebody who knows how
to build a team and knows the problem domain. They have to understand the
toolkits and techniques very, very well and know the math behind them. So
my advice, which is applicable in many places, is to make sure the first hire
is super strong and just really cling to them, because you have no alternative
but to identify people like them. And it doesn't have to be somebody who's
from your field, but somebody who can actually explain and show that they
understand your particular business and what you think. They should not only
be excited about the new technology, but they should also be equally excited
about making the business thrive.
Gutierrez:
Are data science skills from one industry applicable to another
industry?
Radinsky:
Yes. I moved from one industry to another many times, so it's
doable. But, again, each time you move, you've got to work with somebody
who really understands the problem. You have to understand the problem to
be able to apply the tools that you've already seen, and use techniques based
on the data you've investigated.
Gutierrez:
What's a big thing you've changed your mind about using data?
Radinsky:
I think it's that it's not enough to have a statistical background to
understand data and how others see data. The problem that most people
have is mostly about the perception problem. It's more about how humans
perceive it to be correct. I think this was the most surprising thing for me.
