When he assembles a data science team, Scott Nicholson, the chief data scientist at Accretive Health, looks at both the individual and the group dynamic. He wants driven, inquisitive individuals who understand the enterprise mission. And he wants his assembled group to have specialized expertise that can complement each other.
“When you are building out a data science team and you have multiple people with multiple different skill sets, you bring all those data scientists together and then you end up with something that is much greater than those individuals,” Nicholson said. “Because you have people with different backgrounds. Economists. Computational biologists. Astrophysicists. Statisticians. You bring all those people into a room and then you are solving problems in a way you never really thought about before.”
It’s not easy to build such a talent bench, Nicholson said at a recent Predictive Analytics World conference in Boston. In a presentation on the data scientist’s job, he said it is particularly challenging to find talented people who can do everything from asking the right questions to then working with data to find insights and then make those insights actionable.
When he interviews people, Nicholson said he emphasizes these qualities over those who are familiar with Cassandra, the open-source columnar database, and other “big data techniques.” Nicholson’s priorities for data science job candidates include:
Data nerds with investigative zeal. He is not seeking people who have worked at Facebook or other big data-driven companies, but rather those “who think about problems and then are excited about going out and finding data to answer questions.”
“This is not about algorithms. It’s not about industries. It’s just about being curious about data. And they don’t have to have a particular background. As of right now the kinds of people I’m talking to are physicists, electrical engineers, computational biologists, it’s across the board. Because they are just data people who are driven by curiosity,” Nicholson said. “Those are the best data scientists you can find.”
Fluency with datasets and analytical models. Nicholson said his favorite interviews are discussions about using data to solve big problems. “It’s just a conversation of ‘I want to do this, and here’s the data I’m going to use, here’s where I am going to get it, I’m going to use this model, here’s where I am going to put it,’” he said. “For me, that is like table stakes. Let’s start there. That’s the most important thing in a good data scientist. Now we also need people who are in the corner and they got a machine learning computer science degree [and] are technically amazing. And they can implement like no one else. That’s really important. But that work doesn’t’ really go anywhere unless you are thinking about this end to end process.”
Attraction to important problems in business and society. Data science is industry agnostic, Nicholson said. Good data scientists go where the interesting problems are, and where there are datasets to address those problems.
For these reasons, he said he expects more data scientists will join him in the health care industry, where there is more data being generated. With health care making up 18 percent of the U.S. gross domestic product, “if you can improve things by 0.01 percent, number one, that is a huge improvement in patient care, and number two, it is a giant bucket of money.
“If you have this combination of doing good for society with huge financial benefit, and we’re finally at that stage where we have data, and so as data scientists it’s a pretty exciting time and that’s really why I left LinkedIn for health care,” he said.