Having been on both sides of the interviewing process many times, these are 5 tips for interviews that should be helpful to those interested in interviewing for data science roles.

1. When studying, focus on breadth without sacrificing depth

Data science interviews can be tough because depending on what kind of role you're interviewing for, there can be significant breadth AND depth in the types of interview problems you may get asked. You can think of there as two dimensions upon which to gauge this on a spectrum: the company, and the role. Let's take Facebook product analytics and Netflix machine learning as two examples. Facebook has many products and the focus for data science is generally on analytics and experimentation, and hence the focus is on breadth on topics rather than depth. Netflix, by contrast, has one product and most functions are dedicated to optimizing particular aspects of that product (machine learning roles for recommendation for example) and hence the focus is on depth instead. Therefore, for a product analytics role you might get a variety of questions ranging from SQL to probability like the following: a) Assume you have tables on user actions. Write a query to get the active user retention by month. b) Say you roll three dice and observe the sum of the three rolls. What is the probability that the sum of the outcomes is 12, given that the three rolls are different? Whereas on the more technical role (machine learning) and company side you might get more in-depth technical questions such as: a) Say you model the lifetime for a set of customers using an exponential distribution with parameter λ, and you have the lifetime history (in months) of n customers. What is the MLE for λ? b) What is the loss function used in k-means clustering for k clusters and n sample points? Compute the update formula using 1) batch gradient descent, 2) stochastic gradient descent for the cluster mean for cluster k using a learning rate ε.

2. Admit when you don't recall particular concepts

We have all had occasions where a certain concept seemed familiar but not readily available during an interview. A bad thing to do is to pretend like you understand the concept. The best thing to do in this scenario is to admit you don't remember, and let the interviewer guide you. Although this seems obvious, it is very important because a good chunk of the interview process will generally be dedicated to screening for false positives (imagine you have a model simply predicting interview outcomes just with a resume), i.e. cases where the resume is amazing but the candidate is much less so. Having integrity on what you firmly know and don't know is definitely a dealbreaker so it is important to keep this in mind during interviews.

3. Be unafraid of the technical details

Asking technical questions is a way for an interviewer to evaluate 1) whether you understand what you claim to know, and 2) whether you can convey the appropriate information in a clear and concise manner. Often times, these questions are the "make-or-break" points within an interview (especially if you claim to know certain topics but demonstrate very little understanding of questions they would expect you to know, as per the above point). The two qualities being assessed are very important to data scientists because the job revolves around thoughtful and clear data analysis. Therefore, when you have expertise on various technical aspects, definitely be unafraid to show and discuss it with your interviewer. A good interviewer who might disagree with your points will engage in a thoughtful intellectual back-and-forth, and this is definitely not detrimental to the process as long as you are clear with your assumptions and thought processes.

4. Discuss your projects thoroughly

Outside of technical questions, there are often many discussions about the projects you've listed on you resume. Having projects is important because much of data science is about exploration and showing that you have certain demonstrable skills in areas of interest are a great hiring signal. Because of this, hiring managers will often want to get a clear sense of exactly what you did in each project to understand what skills you applied or learned in the process, and how. This means that if you have many projects in groups, you should be ready to explain your significant contribution to each project along with the practical takeaways of each.

5. Understand the product and business side

This applies almost universally across all roles within data science, because irrespective of function, data science will always be about driving product (and hence the business) forward. By understanding the company that you're applying to well, you can get a good sense for both the types of product questions they might ask and where data science fits into each product. For example, at Facebook, the process involving product questions generally revolves around the actual core sets of products (Facebook, Instagram, WhatsApp, etc.) or on hypothetical ones of interest (something newer in the AR/VR space, blockchain, etc.) rather than something tangential. Additionally, the focus is on rapid experimentation (A/B testing, monitoring many metrics, etc.) and thus should give you an idea of what types of product questions they might ask, ranging from measuring success to evaluating tradeoffs during the A/B testing process. This framework has been generally useful across most tech companies in my experience.

Conclusion

Hopefully, this article gave you a couple of useful pointers for interviews. If you are interested in more questions (and answers), make sure to subscribe!