Machine Learning Interviews

Chapter 5. Technical Interview: Coding

In the previous chapters, I walked through the ML interview process and the ML algorithms and model training portions that are part of the technical interview. Technical interviews can ask for much more of candidates beyond ML algorithms, stats knowledge, and model training, though. This chapter will cover one of those pieces, which is the coding interview.

For jobs in machine learning, the kinds of coding that could be asked for will differ among companies and even among teams in the company. For example, when I was interviewing for data scientist and MLE roles, I got the following types of coding questions and tasks:

• Company 1: Python questions related to data manipulation in pandas

• Company 2: Python brainteaser questions (“LeetCode style”) only

• Company 3: Data-related coding questions in SQL and Python Pandas

• Company 4: Take-home coding exercise with a real-life scenario to code

…and so on.

There is a large variance in what companies might ask in a coding round of interviews. From what I’ve seen personally and heard from colleagues working as software engineers and hiring managers of software engineers, the ML coding interviews are less standardized than the technical interviews for software engineering roles. On the bright side, interviewers for some ML roles don’t always ask candidates the most difficult “LeetCode style” questions, aka “LeetCode hards,”¹ since the candidate can be evaluated on their other skills, such as ML algorithm ...