As part of Springboard’s effort to put together a comprehensive, free guide to acing the data science interview, we interviewed Springboard alumni and one of our mentors to get a handle on how they made it past the data science interview. Springboard teaches data science by pairing learners with one-to-one instruction with leading industry experts.


1nirajsheth

Niraj ShethData Analyst at reddit, Springboard graduate

What is advice you’d have for how to ace the data science interview process?

I wish I had studied more fundamental statistics before interviewing. It’s silly, but people often look for whether you are familiar with terms like Type I and Type II errors. Depending on the time you have, I suggest getting a statistics textbook and at least becoming familiar with the terms out there.
I should have probably expected this, but I was surprised how poor we are as an industry in evaluating projects. When I talked about past projects, everyone just cared about interest value (does the analysis say something interesting?) — nobody questioned deeply the methods I used.
You didn’t ask this, but there were also some things I did that I think worked out well. One is to have a live project up somewhere with a neat visualization (i.e. more than a github repo with a readme). It doesn’t have to be fancy–just prove you can build something that works (mine was a fog prediction map, for example). It definitely helps get your foot in the door.
The other thing is to ask for a take-home data set. I don’t know about you, but I’ve found that for myself and other people who don’t have a formal data background, it can be intimidating to work on a data set on the spot; I just hadn’t developed the muscle memory for it yet. However, I knew the right questions to ask, and I could figure out how to answer them if I had a little time, so getting a take-home set let me show what I could do that way.


Saraweinstein2

Sara WeinsteinData Scientist at Boeing Canada-AeroInfo, Springboard graduate

What is advice you’d have for how to ace the data science interview process?

In terms of preparation, I wish I spent more time thinking about analytics strategy. I prepped hard on stats, probability, ML, python/R…all the technical stuff, but was nearly caught off guard by a straightforward question about how I’d approach a particular problem given a specified data set. My answer wasn’t as confident as I would have liked. I’d been so focused on the “hard” stuff that I hadn’t thought that much about higher-level analytics methods & strategies.

What surprised me and what I found difficult:

How long the process took. I knew to expect several interviews, and in fact had three. With nearly a week between each, plus waiting for my background check to clear, the process from first contact to firm offer took a month. It was stressful to say the least. Staying positive, confident, and prepared for a whole month was challenging. It would have been much easier to bear if I’d known in advance that it would take that long. For others facing a lengthy multi-interview hiring process: meditation is your friend. It helped me sleep at night, and I used the techniques right before interviews to channel calm and confidence.


sdrjansantic3Sdrjan Santic Data Scientist at FeedzaiData Science Mentor at Springboard

What is some advice you’d have for how people can ace the data science interview process? What were some of the toughest questions?

The most important thing, in my opinion, is understanding how the major supervised and unsupervised algorithms work and being able to explain them in an intuitive way. A good command of Data Science terminology is crucial. Candidates should also have a thorough knowledge of relevant accuracy metrics, as well as the various approaches to evaluation (train/test, ROC curves, cross-validation). The tougher questions would
relate to these same affairs, but with having to break out the math on a whiteboard.

How did your interview process go?

Luckily, very smoothly! Most of my interviews had a feeling of being a conversation between peers, so I didn’t find them very stressful. The companies I interviewed with moved very quickly (one round a week), which helped streamlined the process. I was also very impressed as to how most companies that turned me down gave me very honest feedback as to why!

What were some of the factors for you in choosing your current job?

Primarily, it was the opportunity to use a technical toolset and solve problems I hadn’t solved before. My previous role was very focused on just building models. The data was already completely cleaned and pre-processed, and the exploratory work was done using a commercial GUI-based tool. I felt that my data-wrangling and command line edge was being dulled slowly and jumped at the opportunity to work in an environment where I’ll be able to “get my hands dirty” once more!

 

Like this article? Subscribe to our weekly newsletter to never miss out!

Image: Sebastian Sikora

Previous post

Data Around The World - Part II: First Data From The Itility Hackathon

Next post

"Smart Data Is The Future of FinTech" - Speaker Spotlight: Q&A with Patrick Koeck - Data Natives Berlin 2016