News & Ideas

Data Science Job Candidates – What I Look for As a Hiring Manager

Kamil Grajski, TROVE’s Managing Director of Data Science, reveals what he is really looking for in a crowded candidate-pool.

In recent weeks, I’ve interacted one-on-one with nearly 100 Data Science job candidates for a handful of open positions at TROVE – a growing and successful predictive data analytics start-up where I am the Managing Director of Data Science.

I started the search thinking that I was looking for a “needle in the haystack” and proceeded in typical fashion choosing from a standard set of questions to drive candidate interviews. (1) What is your educational background? (2) What is your work experience? (3) What kinds of data and with which industries have you worked? (4) How proficient are you in R? Python? C++/Java? (5) Are there code and technical report samples that you can share? (6) Has your work resulted in the commercial launch of the product? 

In the not too distant past, these questions were enough to gain a rough appreciation for a candidate’s experience, skills and potential fit with the company. It was unusual to find candidates who could broadly and deeply range across these topics.

What I discovered, to my surprise, is that today nearly every candidate “made the grade”! I needed new criteria to have any hope of fairly and accurately ranking candidates.

I landed on a set of four new questions that consistently led me to discover what it was that I was really looking for– the Scientist within the Data Scientist:

  • Which of your skills are ready to be put to use as a professional Day One? Candidates, perhaps in a bid to catch the eye of machine learning resume readers, load their resumes with long lists of skills. This is not helpful. Do your homework on the company. Anticipate a likely Day One dataset and be prepared to identify challenges in handling such data and describe applicable statistical modeling techniques.
  • Do you have an investigator’s mindset? Candidates routinely analyze coursework datasets through a laundry list of modeling methods. Have you demonstrated a deep understanding of the core principles underlying a given statistical method? What is your process for formulating and testing hypotheses about the data?
  • Are you a lifelong learner? This question is designed for candidates who have been working for more than a few years as professional Data Scientists. What statistical computing environment have you used in the past year for the first time? What statistical or machine learning method have you applied for the first time in the past year? Why?
  • Do you respect the data? Leo Breiman, a luminary in machine learning through pioneering work on classification and regression trees, bagging and random forests, would begin the first lecture of his year-long multivariate statistics course at UC Berkeley by writing on the board, “Respect the Data.” Why do you think he did that?

Armed with these additional questions, I was able to identify candidates who have what I need most on my teams: a passion for discovery combined with a rigorous scientific approach to the insightful application of modern statistical analysis and machine learning methods.


Fill out my online form.

Thank you.
We'll be in touch soon.