Database Reference
In-Depth Information
playbook for things, so you really have to be able to incorporate whatever is
new and interesting at the time. Those are the big things we look for when
we hire: someone who is smart, who knows the basics of math, statistics, and
programming, who can learn really quickly and incorporate new technologies,
and who has knowledge that we do not already have.
Gutierrez:
What advice would you have for people who are hiring their first
data scientist?
Hu:
When I was hired as the first data scientist at Next Big Sound, there were
a lot of people on the team who were already very data-savvy and maybe did
not have the bandwidth or the specific statistical background to do some of
the work. And so they were capable of assessing the data science candidate's
quality. If you do not have that background, I think it would be tough to really
discern between different qualities of data scientists, because there is really a
huge spectrum of people out there.
If you do not have some institutional knowledge, it would be difficult to tell
who is qualified or not. This is especially true because there are not very
established programs in most universities. That said, there are a lot of new
programs that are just now coming into play, so this could change in the future.
Currently there are a lot of people learning from online courses and from
just doing it themselves, which I think is great. What I look for when hiring
is people who have done projects, specifically who have delivered concrete
insights using machine learning and who know how to communicate that.
Gutierrez:
What does delivering concrete insights mean to you?
Hu:
At the end of the day, if you are not changing the behavior of the cus-
tomer or the industry that you are in, then it is hard to assess the value of
your insights or your work. So that is what I think about every day in the
research that I am doing. I ask myself: is the new product capability that I am
thinking of actually going to change the workflow of someone in the music
industry? Is the research going to change how they find artists, or how they
market their artists, or how they decide to tour or release albums, or some-
thing that actually affects their decisions? This line of questions applies to any
industry and to any field of data science.
Volunteering at the DataKind Data Dive recently was really illustrative of that
because so much of what we are trying to do there is to affect nonprofits that
really touch thousands, if not millions, of people. We were working with places
like the UN, the World Bank, and Amnesty International. I was helping lead the
Amnesty project, and we were able accomplish a lot in a short amount of time,
because they were there to provide immediate feedback on what specific
products or insights would actually affect their day-to-day. Having the NGO
representatives and the people who have that industry-specific knowledge
working closely with the data scientists the whole weekend was, I think, what
made it a lot more impactful and successful. Having that quick feedback is key.
