Database Reference
In-Depth Information
These days in data science, I also see that best practices are being shared.
Universities have set up courses. Top universities in the US are now giving
you the opportunity to get a data science degree. And when you graduate
as a data scientist, you will come into an environment where data integra-
tion has already been solved by companies like Planet OS, Trifacta, Tamr, and
where feature engineering has been solved by companies like SparkBeyond.
Furthermore, nonparametric search methods, which James Burke and others
have talked about, are going to allow computers to give you ideas on which
features are the best. Other things that are going to be partially solved are
visualizations tools like D3.js and Tableau, which are making it easy for anyone
to make great visualizations with great insight. Also helpful will be the exten-
sibility of the language that is used for developing the models. For instance,
probabilistic programming will allow you to have variables that represent
probability distributions, so you can start to develop in completely new ways,
as compared to today with the old languages.
In the future, the tools around you are going to be such that you can think
more about the business problem and less about the technical side. You'll have
to think less about how it scales and how the data comes together, because it
will be more and more automated. You as a data scientist will need to learn
to understand the business you are in much better, as well as learn to inter-
act with other domains and with people from other disciplines. I think it is
Chevron who has executives paired with data scientists so that when there
is any big investment decision to be made, they work together to select the
best course of action.
Similarly, these days, you are starting to see chief analytical officers next to
chief data officers. The chief analytical officer represents the change that any
major decision throughout the organization is now getting made by everyone
in the organization, with data scientists and machine learning methods being
the means and tools to make those better decisions. You are now seeing data
scientists being planted into product teams, instead of being an isolated island
or a group of experts throwing wise words at other groups. I think eventually
we'll have something similar to the Agile Manifesto and agile methodology for
data science—agile data science or whatever else you want to call it—that
will bring some old methods and ways for collaborating and working in this
field. And then suddenly, the focus will shift from methods and tools to actual
end results and the delivery of those results.
Gutierrez:
What advice would you give to someone starting out?
Karpištšenko:
Though somewhat generic advice, I believe you should trust
yourself and follow your passion. I think it's easy to get distracted by the
news in the media and the expectations presented by the media and choose
a direction that you didn't want to go. So when it comes to data science, you
should look at it as a starting point for your career. Having this background
will be beneficial in anything you do. Having an ability to create software and
