Types of Machine Learning Interviews and how to ace them
A detailed guideline on different categories of machine learning interviews and how to prepare for them.
In the Spring of ‘21, I started applying for jobs. I couldn’t find one concise article on what are the different aspects that I will be interviewed on. In the initial interview series, at times I was clueless. I had to learn it the hard way that no matter how good a researcher or ML engineer you are when it comes to interviews that are not enough. You must prepare for it! Before you can even start preparing, you need to know what are the different areas that you will be tested on.
In this article, I will discuss the different types of Machine learning interviews based on the 58 hours of interviews I have had earlier this year. The pie chart below summarizes the different categories of Machine Learning Engineer/Data scientist interviews.
We will focus on screening, coding, machine learning, case study, and system design.
1. Screening
This interview is rather casual, and most often the first step into the series of interviews. Its normally conducted by a recruiter or a hiring manager. The main purpose of this interview is to give the candidate an idea of the company, job description, and responsibilities. The candidate is also asked about his/her background in a casual setting to see if the candidate’s field of interest aligns with the job.
Nature: Non-technical
Mode: Normally a Phone Call
Duration: 15–20 mins
Preparation: You should be able to explain your experience in a few mins.
2. Coding
Coding/programming is an important part of the machine learning interviews, and more often than not are used to filter out candidates before moving them forward to the machine learning-related interviews. Good programming skills are required to perform in such interviews. Coding interviews usually last about 45–60 mins and consist of two questions. The interviewer explains the problem and expects the candidate to solve it in optimum time and space complexity.
Preparation resources:
Preparation of such interviews requires a good understanding of data structures, time-space complexities, the ability to understand the problem, and good time management skills. Following are some good resources
3. Machine Learning
Machine learning interviews assess your knowledge related to Machine Learning. Based on the job requirements, the topics can include supervised learning, unsupervised learning, reinforcement learning, convolutional neural networks, recurrent neural networks, generative adversarial networks, natural language processing, etc.
Preparation resources:
Make sure you go through the job requirements and identify specific topics of ML that are required. If nothing specific is mentioned, then you can focus on the basics.
a. An in-depth course on ML: If you have not formally taken a course on ML during your studies, I would strongly suggest you take one online. Among the various ones available online, I would recommend Andrew Ng’s Machine Learning course offered by Stanford available on Coursera.
b. Refresher Articles: If you have already taken such a course, then going through a few refresher articles on ML will be really helpful for you to create an in-depth understanding of ML topics. A small list (10) of such articles can be found below.
c. Sample interview questions: Getting an idea of what usually is asked in ML interviews helps gauge your preparation. Such example questions can be found below. Make sure you go through them AFTER you have done your preparation.
d. Cheatsheets: It's always good to have cheatsheets that you can go through the night before your interview. Below is the link to an ML cheatsheet
4. Case Study:
These are usually open-ended questions aimed at analyzing the candidate's ability to have meaningful ML discussions on the overall project management and project acumen. An open-ended problem is presented by the interviewer, such as “How would you improve google maps?”. Such open-ended questions can result in a very chaotic answer if not approached properly.
Preparation resources:
A good answer template goes a long way in approaching such questions. Try to organize your answer using the following template
- Listen to the question
- Describe the product and its mission
- Ask clarifying questions
- State your assumptions
- Identify the pain points
- Identify solutions to the pain points
- Compare solutions through the table
- Discuss the KPI
- Summarize
Following is a good resource in understanding case study questions in depth
Some examples of such open-ended questions can be found below
5. System Design:
These interviews assess the candidate’s ability to design an end-to-end scalable system for the underlying problem. Most of the engineers are so focused on a bug that they forget or at times even fail to realize the bigger picture. A system design interview requires knowledge of various components that come together to create a scalable solution to the problem. These components include front-end design, load-balancer, database sharding, caching, proxies, SQL, API, etc. A good understanding of these topics helps design an efficient and scalable end-to-end system.
Preparation respources
- Components and concepts:
2. Examples Questions and Solutions
Summary:
Machine learning interviews these days are much more than just questions on basic ML topics. They also include open-ended questions, case studies, coding, system design, etc. Understanding different categories of machine learning interviews can help the candidate be informed and prepare accordingly. In this article, we went through in detail on five most common ML interview categories and how to prepare for them.
If this article was helpful to you or you want to learn more about Machine Learning and Data Science, follow Aqeel Anwar, or connect with me on LinkedIn or Twitter. You can also subscribe to my mailing list.
