Database Reference
In-Depth Information
Gutierrez:
What are the challenges going forward?
Karpištšenko:
Now as businesses rely on our services and software, we will
have to think about all of the service level agreement aspects of our work.
We will also have to work out customer onboarding, support, and mainte-
nance. We'll have to make sure that our roadmap priorities reflect what the
most important customers require while also making sure that these deci-
sions don't paint us into a corner. This way, the business has the ability to
grow and scale beyond our wildest dreams. The challenges are going to be
around scaling and operational excellence for a while, rather than innovation
and development as it used to be.
Gutierrez:
What do you look for in other people's work?
Karpištšenko:
For work that presents novelty, I want to know how it com-
pares to something else that I know. If someone presents a new technology or
a new method, I need to know some benchmarks or some baselines against
which they are comparing. Without it, I just disregard it because I don't have
time to do the comparison myself. With things where someone presents a
success story or a use case, I look at how they approached the problem
holistically and whether they are highlighting similar problems and challenges
I myself am familiar with.
In cases when I look at work related to where I see the industry moving in the
future—toward sub-second analytics, toward exascale data sets, and toward
higher involvement of nontechnical people in analytics—in these cases, I try
to get inspired and to remember the lessons learned or the models they have
used to succeed, so that, when the opportunity presents itself to me, I have
something in my backpack to pull out and start iterating from.
Gutierrez:
What do you think the future of data science looks like?
Karpištšenko:
Automated. To me, I see many similarities with what hap-
pened in the software industry in the 2000s, when the Agile Manifesto came
out, as compared to what was there before, which was this rational unified
process everyone talked about.
I see many similarities with what's going on in the data science community
right now. In the software industry, there used to be a lot of focus on how
it was difficult to deliver on time, how there are so many uncertainties, and
how so many software projects failed. Then a new method, inspired, again, by
the car manufacturing industry and by old production companies, brought
new ways of working on software development projects. The Agile Manifesto
actually made it so that these days everyone knows how to deliver software
projects. If someone doesn't deliver, then it's much easier to understand what
went wrong and why they were unable to ship. Of course, it's a different story
if it is a highly innovative, high-risk, high-tech project.
