As part of the release of Prolego's new research report, "Winning the War for Data Science Talent" we are digging deeper into the tactics of building a world-class data science team. Recommendation #7 of the report:

Optimize your recruiting process for candidates who are curious and are good communicators. Evaluate hard skills based on past work. 

Justin Pounders, Prolego's VP of Data Science and Kevin Dewalt, Prolego's CEO recently discussed Prolego's interview process.

‍

Here is a summary of our conversation. 

The job of a data scientist at Prolego

Unfortunately the job title of “data scientist” is an overloaded term whose meaning varies widely. At Prolego a data scientist’s job is to develop new ways of solving complex business problems. 

Our data scientists evaluate our customers’ businesses to learn ways that data science might help them work more efficiently, constantly review new and emerging techniques from the research community, combine techniques from multiple domains, and evaluate the effectiveness of them on our customers’ data and operations.

In most situations our team is starting from a blank piece of paper where even the problem isn’t well defined. We rarely encounter situations where a playbook or checklist applies. Our work looks nothing like a Kaggle contest where the objectives are clearly defined and the data is available.

Our interview process

Resume reviews

We start by looking at someone’s background to see if they have demonstrated experience applying first-principles reasoning to solve a new problem.

A major red flag is a listing of certificates or academic credentials without corresponding work. We ask ourselves, “what has this person done?”, not “what does this person know?” The ideal candidate has demonstrated curiosity about a particular problem and invested time exploring it further. A github repo and a series of blog posts that demonstrates curiosity and competence is vastly more important than a list of credentials.

Initial screening interviews

Unfortunately there isn’t a packaged set of interview questions. In an initial screening conversation we did deeper into what they have done. This conversation also gives us the opportunity to evaluate communication skills. We need someone who can communicate effectively with everyone from CEOs, peers, customers, and other engineers. A candidate who can engage in a smart, curious conversation with us usually has these skills. Someone who continuously talks or who doesn’t ask good questions usually doesn’t.

Finally, Justin’s favorite question is, “How do you learn something new?” The ideal candidate learns by testing and trying new things, not simply reading about them.

Detailed interview process

Candidates that get through the initial screening process spend more time with the rest of the team to evaluate harder technical skills and cultural fit. We evaluate hard skills by digging in deeper to what they have done through conversations and code review. We usually ask them to walk us through a past project.

We also provide them with a problem or set of code and ask for their point of view and opinions about solving it. We’re primarily looking for their thinking and analytical process.

Candidate selection

Finally we finish up with interviews with the rest of the team to mutually evaluate whether it is the right fit. This step isn’t much different from interviewing any other candidate.

Interviewing is only part of the strategy

I hope this conversation helps you refine your data science interview process. However, the right process won’t help if you don’t have a plan for becoming an employer of choice and attracting the right candidates.

We cover these topics in detail in How to Win the War for Data Science Talent. Get your free copy today.