A successful interview for a data scientist position relies on the ability to effectively communicate and demonstrate your combined experiences and skills.

With high demand comes high competition for those seeking data scientist roles. Interviews are still an important part of the hiring process and can present a problem for even the most qualified candidate.

Doing background research on the company, going over neglected skills and preparing some anecdotes can help ease the tension for candidates.

To further prepare, it is necessary to understand where interviewers are coming from and what they are looking for in an employee. Here are some of the major factors interviewers are looking for in candidates and some procedures companies will run you through during the interview process.

Preparing for the interview

As with most things, laying groundwork is crucial to the success of an interview. For a highly technical and difficult position such as data scientist, this is especially true. Going over skills, recalling solutions to past problems and analyzing each step can help candidates prepare for the process.

“It’s the drive to understand and solve decision-related problems that marks a great potential candidate,” said Juan José López Murphy, technical director and data science practice lead at Globant, an IT and software development organization.

For the interviewer, the ideal candidate has the right knowledge and is able to communicate it. This means those seeking data scientist positions need to prepare for technical questions and cultural questions. With a team mentality present in companies, data scientists are expected to be willing and able to share their knowledge and skills.

Preparing for these questions is important because an inability to communicate skills or an unwillingness to work as a part of a team can hinder your chances.

“Tell us a business challenge as a story, using an example from your past experiences,” said Steve Tycast, director of data and analytics at AIM Consulting Group. Even projects that couldn’t deliver the results needed can be seen as a positive for a candidate to discuss.

What are interviewers looking for?

When it comes to interviewing preparation, technical savviness, and communication skills are of equal importance.

“It is expected that a candidate will be good at their core discipline, that they know Python or R or Matlab,” said Gus Walker, senior director of product management at Veritone. “It is important that they demonstrate they have the skills to develop AI models, but it’s equally important that they are willing to share that knowledge, that they are curious about how we develop our solutions beyond their specific contributions.”

This innate or learned curiosity can give candidates the edge in interviews for their desired position. Being familiar and skilled with Python or R is still important, but companies ask for more from their data scientists.

Be prepared to handle questions that range from highly technical to general when it comes to coding languages and learned skills. It is important to be familiar with basic statistical concepts and their interpretations, as well as some common algorithms.

But these will most likely be supplemented by questions that seek to prove something about the candidate. Interviewers want to know their candidate has a drive to learn, the ability to communicate and can understand how to take an idea all the way through to production.

“I like to ask [for] a scenario where the merits, benefits and cons of different approaches can be discussed,” Murphy said. “It sets the stage to discuss the candidate’s rationale and not so much the memory recall of a textbook formula.”

More important than an eidetic memory to interviewers is proof that a candidate has a curiosity about the subjects at hand and is willing and able to work through problems without reading an answer as if it were from a textbook. Anecdotes of past experiences cement your experience in the mind of interviewers. They are looking for humans, not robots.

Example questions

• How do you measure model performance across learning strategies?
• What methods would you use to help verify the completeness of the data received through an ETL (extract, transform, and load)/ELT (extract, load and transform) pipeline?
• If a model you have developed performs at 90% accuracy, what do you need to know to interpret whether this is good or not? How would you prove this?
• Explain to me, using an example from past experience, what is bias and variance tradeoff?

Approaching answers

When it comes to answering these questions, it is important for applicants to be able to demonstrate their skills.

“We are looking for answers that show that the applicant not only understands the problem, but understands and can explain why the solution they provide is the best from the perspective of accuracy, efficiency, simplicity and maintainability,” Walker said. “We also look for answers that demonstrate creative problem-solving.”

Interviewees should walk the interviewer through the thoughts that led to the answer. Being able to share these processes and demonstrate problem-solving abilities is crucial to the role and, therefore, crucial to the interview.

“Almost every answer should be stated in a relative form to the matter at hand,” Murphy said. “State your understanding of the problem, state the rationale of the kind of answer you seek, then the first steps you would take to solve it and how you would measure or understand what’s working and what’s not working and how that informs your next steps.”