Ace your next Data Science interview | by Lan Chu | Jul, 2022

All the real stuff with questions in great detail

In this post, I will talk about the data science interview process and the questions that were asked based on my experience. Many people are intimidated by interviewing, however, notice that it is a candidate’s market right now and this process can be fun if you know what to expect and how to prepare. It is important to keep in mind that depending on the companies, the process differs. Even within the same companies, different interviewers may have different approaches. Not all interviewees get asked the same questions. But in general, there are a few similarities that you should expect to see in most of the Data Science jobs that you apply for.

The goal of this introductory interview is to understand better your situation and a bit of your study and work experience. Nothing technical at this point. This conversation is rather short and usually lasts around 15–30 minutes.

In my experience, this round is not too much about the technicality, unless the manager has heavy quantitative/technical background. They will focus on understanding your goals and expectations, the reasons behind you applying for the job, and how you fit in the position, the team, and the company culture. Occasionally, they will ask questions for a broad assessment of your relevant skills. Some questions I got asked include:

– What do you know about the company.

This is an ancient question, but it does happen to me more than once. It is probably not in your best interest showing up and being clueless about the company you want to work for. Plus, it is useful to know about the company and the values it stands for. Recruiters get excited about your passion for the company’s mission. For me, it is also important that the company I am going to work for follow certain values that I appreciate. Well, at least on paper.

– Why do you apply? What is your expectation of the role?

For this kind of question, think about how you want to contribute to the team and how the job will benefit your learning and your career path. To be convincing, you should have a clear idea of what you want to pursue in your career. And even if you don’t, pretend to!

One useful tip is to think about the pain of the company — the reason why they are hiring you. They do not hire you for nothing, it serves a purpose and a need that is not yet being met. This again requires doing research ahead of time and making an educated guess. With that, you will appear to be more of the right fit because you understand what they need and what you can do!

– Explain your previous relevant experience from a functional side.

Easy peasy lemon squeezy. Expect them to ask questions about the past projects you have worked on. The focus, in general, will be on the functional side. You should not go technical here. They will get bored.

– How you collaborate with colleagues.

Teamwork is crucial for any project. It is a high chance that you will work with colleagues who come from different backgrounds, culture,s and beliefs. Think about a time that you face a difficult situation at work, either challenging deadlines or conflicts at work, and how you solve that.

This interview is rather a technical, math, and concept interview. I think there is no fixed set of questions for this. The conversation will evolve depending on how you explain and answer things and the experience/knowledge of the interviewers on a certain topic.

– Describe functionally and technically a DS project you worked on (they will pick one of those in your CV)

In one interview, I got asked to explain the Image Recognition project I mentioned in my resume.

• The size of the data and any potential data quality issues. They want to know if you do proper Exploratory Data Analysis
• Gently dive into the math behind the convolutional neural network 😐.

And in another interview, the interviewer dived into a time series forecasting model, again a project in my CV. Many questions were asked, you can expect some questions like the following:

• Which metrics did I use to measure model performance (MAE, RMSE, AICc etc…). Why did I use one metric over others?
• How did I do cross-validation. So, explain the idea behind cross-validation
• Explain the chosen feature selection techniques
• One of the models I used was Random Forest Regressor. List 3 hyper-parameters of Random Forest and how to do hyper-parameter tuning ( e.g Gridsearch Sklearn or any other framework)
• Another model I experimented with was ARIMA. One interviewer asked the model is basically auto-regressive. Considering the context, do you think that is the reason why your model performed badly.

– The ethical issue in Data Science

The interviewer explained a real business situation and asked whether the way the company dealt with it was ethical. You must be wise here. You want to be true to yourself, but you would not want to offend your potential employer either. It requires understanding the context and some ethics and fairness concepts in Data Science. Try to listen and analyze the situation, and ask questions when you are not sure.

But if you can’t seem to provide an answer, don’t worry. I can save you. A perfectly fine answer could be: the line between ethical and unethical issues sometimes is so thin and without further context, it is difficult to decide 😉

– I have data but I do not have a business case. What can I do?

In real-world scenarios, many times stakeholders will come to you with the data and ask you to propose some project ideas/initiatives that are beneficial for them. It is far from easy but one thing you can always do is to first explore the data, visualize it, and create a dashboard to provide insights. Link it with the topics that are relevant to the stakeholders.

Once you passed the technical interview, the next step can be a coding assignment. You work on the coding exercise within a limited time period. The result and the code will be reviewed by the responsible person(s) and they will decide if they want to move forward with you.

The coding assignment can differ among companies. It could be, literally, a coding exercise, in which you are asked to solve a real-world business request, i.e. join different data sets and report on the number of unique customers. It could also be working on a business case. You are given a data set and a business question to solve. With this assignment, the hiring company wants to test not only your modeling and programming skills but also your overall approach when you face a problem like this.

The good thing about this exercise is that it is usually a take-home assignment. Lock yourself in a room and work on it. You have all the resources you need, the book, the internet. Google and Stack Overflow are there.

I think the reason why companies allow you to do this at home is that, actually, in our daily work, we google too. It is perfectly ok to consult StackOverflow or other resources and digest those code solutions. It’s a normal learning process. As long as you know how to google (yes!), are willing to learn, and are able to connect the dots, they are satisfied. Similarly, the company is not so worried whether you ask someone else to help with this exercise. You discuss and get help from colleagues anyway. And they can always ask questions during the code review/presentation session to check your understanding of your own code and how you approach the problem.

Because they want to have an overview of your approach, make sure you understand the business question, the data, and that the methodology you use fits the problem you want to solve. You will be asked to send your notebook (probably through GoogleColab or any similar platform) and a ppt presentation. If possible, make an appealing slide deck 😊 It helps to tell a story, and people like it. Trust me, you may do this a lot in your job!

Awesome! You have been selected to present your results. We are almost there. The goal of this is to probe for more details on your approach, whether the solution provided fits and the hiring company wants to test your ability to speak to business stakeholders.

You will be given some time to present your work and questions will follow. Make sure you understand every little thing you put in your notebook and slide. You should expect to receive some questions on your approach (i.e classification or regression), your feature engineering and selection, the model that you choose, the model evaluation and any potential pitfalls.

Usually, you present this to a team of Scientists and Business colleagues, so speak the language of the latter and make sure your story is relevant to them. The data scientist interviewers can go technical if they see the need. I think I went a bit too much on explaining Precision and Recall in my presentation (🙈) You could try to practice in advance with a friend to get feedback.

All the real stuff with questions in great detail

In this post, I will talk about the data science interview process and the questions that were asked based on my experience. Many people are intimidated by interviewing, however, notice that it is a candidate’s market right now and this process can be fun if you know what to expect and how to prepare. It is important to keep in mind that depending on the companies, the process differs. Even within the same companies, different interviewers may have different approaches. Not all interviewees get asked the same questions. But in general, there are a few similarities that you should expect to see in most of the Data Science jobs that you apply for.

The goal of this introductory interview is to understand better your situation and a bit of your study and work experience. Nothing technical at this point. This conversation is rather short and usually lasts around 15–30 minutes.

In my experience, this round is not too much about the technicality, unless the manager has heavy quantitative/technical background. They will focus on understanding your goals and expectations, the reasons behind you applying for the job, and how you fit in the position, the team, and the company culture. Occasionally, they will ask questions for a broad assessment of your relevant skills. Some questions I got asked include:

– What do you know about the company.

This is an ancient question, but it does happen to me more than once. It is probably not in your best interest showing up and being clueless about the company you want to work for. Plus, it is useful to know about the company and the values it stands for. Recruiters get excited about your passion for the company’s mission. For me, it is also important that the company I am going to work for follow certain values that I appreciate. Well, at least on paper.

– Why do you apply? What is your expectation of the role?

For this kind of question, think about how you want to contribute to the team and how the job will benefit your learning and your career path. To be convincing, you should have a clear idea of what you want to pursue in your career. And even if you don’t, pretend to!

One useful tip is to think about the pain of the company — the reason why they are hiring you. They do not hire you for nothing, it serves a purpose and a need that is not yet being met. This again requires doing research ahead of time and making an educated guess. With that, you will appear to be more of the right fit because you understand what they need and what you can do!

– Explain your previous relevant experience from a functional side.

Easy peasy lemon squeezy. Expect them to ask questions about the past projects you have worked on. The focus, in general, will be on the functional side. You should not go technical here. They will get bored.

– How you collaborate with colleagues.

Teamwork is crucial for any project. It is a high chance that you will work with colleagues who come from different backgrounds, culture,s and beliefs. Think about a time that you face a difficult situation at work, either challenging deadlines or conflicts at work, and how you solve that.

This interview is rather a technical, math, and concept interview. I think there is no fixed set of questions for this. The conversation will evolve depending on how you explain and answer things and the experience/knowledge of the interviewers on a certain topic.

– Describe functionally and technically a DS project you worked on (they will pick one of those in your CV)

In one interview, I got asked to explain the Image Recognition project I mentioned in my resume.

• The size of the data and any potential data quality issues. They want to know if you do proper Exploratory Data Analysis
• Gently dive into the math behind the convolutional neural network 😐.

And in another interview, the interviewer dived into a time series forecasting model, again a project in my CV. Many questions were asked, you can expect some questions like the following:

• Which metrics did I use to measure model performance (MAE, RMSE, AICc etc…). Why did I use one metric over others?
• How did I do cross-validation. So, explain the idea behind cross-validation
• Explain the chosen feature selection techniques
• One of the models I used was Random Forest Regressor. List 3 hyper-parameters of Random Forest and how to do hyper-parameter tuning ( e.g Gridsearch Sklearn or any other framework)
• Another model I experimented with was ARIMA. One interviewer asked the model is basically auto-regressive. Considering the context, do you think that is the reason why your model performed badly.

– The ethical issue in Data Science

The interviewer explained a real business situation and asked whether the way the company dealt with it was ethical. You must be wise here. You want to be true to yourself, but you would not want to offend your potential employer either. It requires understanding the context and some ethics and fairness concepts in Data Science. Try to listen and analyze the situation, and ask questions when you are not sure.

But if you can’t seem to provide an answer, don’t worry. I can save you. A perfectly fine answer could be: the line between ethical and unethical issues sometimes is so thin and without further context, it is difficult to decide 😉

– I have data but I do not have a business case. What can I do?

In real-world scenarios, many times stakeholders will come to you with the data and ask you to propose some project ideas/initiatives that are beneficial for them. It is far from easy but one thing you can always do is to first explore the data, visualize it, and create a dashboard to provide insights. Link it with the topics that are relevant to the stakeholders.

Once you passed the technical interview, the next step can be a coding assignment. You work on the coding exercise within a limited time period. The result and the code will be reviewed by the responsible person(s) and they will decide if they want to move forward with you.

The coding assignment can differ among companies. It could be, literally, a coding exercise, in which you are asked to solve a real-world business request, i.e. join different data sets and report on the number of unique customers. It could also be working on a business case. You are given a data set and a business question to solve. With this assignment, the hiring company wants to test not only your modeling and programming skills but also your overall approach when you face a problem like this.

The good thing about this exercise is that it is usually a take-home assignment. Lock yourself in a room and work on it. You have all the resources you need, the book, the internet. Google and Stack Overflow are there.

I think the reason why companies allow you to do this at home is that, actually, in our daily work, we google too. It is perfectly ok to consult StackOverflow or other resources and digest those code solutions. It’s a normal learning process. As long as you know how to google (yes!), are willing to learn, and are able to connect the dots, they are satisfied. Similarly, the company is not so worried whether you ask someone else to help with this exercise. You discuss and get help from colleagues anyway. And they can always ask questions during the code review/presentation session to check your understanding of your own code and how you approach the problem.

Because they want to have an overview of your approach, make sure you understand the business question, the data, and that the methodology you use fits the problem you want to solve. You will be asked to send your notebook (probably through GoogleColab or any similar platform) and a ppt presentation. If possible, make an appealing slide deck 😊 It helps to tell a story, and people like it. Trust me, you may do this a lot in your job!

Awesome! You have been selected to present your results. We are almost there. The goal of this is to probe for more details on your approach, whether the solution provided fits and the hiring company wants to test your ability to speak to business stakeholders.

You will be given some time to present your work and questions will follow. Make sure you understand every little thing you put in your notebook and slide. You should expect to receive some questions on your approach (i.e classification or regression), your feature engineering and selection, the model that you choose, the model evaluation and any potential pitfalls.

Usually, you present this to a team of Scientists and Business colleagues, so speak the language of the latter and make sure your story is relevant to them. The data scientist interviewers can go technical if they see the need. I think I went a bit too much on explaining Precision and Recall in my presentation (🙈) You could try to practice in advance with a friend to get feedback.