Cracking The Data Science Interview in Australia!
You're not the only one that finds data science interviews hard. That was me a few years ago until I learned a few things to get through it. Let me tell you everything you need to know to crack the data science interview.
So you've already applied for a data science position and you've been invited to interview. In summary, there are three stages to a data science interview process. The first is the screening interview. The second is the technical interview and the third stage is the on-site interview or in-person interviews.
Screening Interview
This stage is the easiest part of the entire interview process. This is simply a phone chat with either the recruiter or a data scientist from the company you're applying to. The goal of this screening is to figure out whether you're the right person to be applying for this type of job and get a general sense of your experience based on what you say in your resume. A basic check as to whether or not you're the person you say you are on your resume. It's a shallow culture type fit and don't be surprised if they're throwing some basic machine learning questions in there, such as what is a validation set or what is regularisation. So be sure to brush up on your basics before you do a screening interview.
Technical Interview
So once you get past the screening interview, you'll most likely be invited to do a technical test or a technical interview. As being a data scientist is a fairly technical role the companies want to know that you have a good foundation in coding and data science this stage can vary significantly in difficulty based on whether you're applying for a junior, mid-level, or senior role. And it's usually done in two ways.
- Online Data Science Quiz
- Take-Home Data Science Assignment
Online Data Science Quiz
The online data science quiz will usually be a timed multiple-choice style question or it can be an online console test that requires you to write some code. Almost all of these tests will require you to answer questions about your understanding of SQL, Python, statistics, and data science.
These tests will be administered in some sort of an online platform such as HackerRank, Testgorilla, TestDome or Alooba. It'll be a mix of theoretical and practical questions. So let's go through some examples across all of those four categories I mentioned.
SQL
So an example of an SQL test could be the following. Given the following data definition, write an SQL query to find all the students who had a test score greater than 80 and whose first name is not John.
TABLE students
id INTEGER PRIMARY KEY,
firstName VARCHAR(30) NOT NULL,
lastName VARCHAR(30) NOT NULL,
testScore INTEGERThe solution is the following.
select *
from students
where testScore > 80
and firstName != 'John';The difficulty of the SQL-related question will vary based on the level you apply for.
Python
Here are some examples of Python questions.
- What is a lambda in Python? (More on Python syn
- How do you filter based on a particular column value in a data frame? (Pandas related questions)
For data science interviews, you will be asked python questions related to data manipulation.
Statistics
Example of statistics question
- Give examples of data that are not normally distributed
And in this case, the answer would be categorical data like the education level of a person or data that adheres to exponential distributions like the amount of time in months that a car battery lasts.
Data Science
Examples of data science questions
- What is a random forest? How does it defer from logistic regression?
And there's a whole variety of questions like that that I found online that helped me greatly to prepare for data science interviews. You can find these in my useful data science resources page.
Alternatively, instead of an online data quiz, you might actually have this technical interview with a data scientist at the company. But the types of questions they ask will be very similar.
Take-Home Data Science Assignment
Let's move on to the second type of technical test which is the take-home data science assignment. From my experience, this is usually a Kaggle style question that requires you to do the working and present it in a Jupyter style notebook and you usually have around two to three days to complete something like this or you might actually get this assignment online and you might have a couple of hours to do it.
An example of this could be a Kaggle challenge which requires you to predict sale prices of houses. They will give you a data set of 100 000 houses with 80 of 90 features and the task will be to create a model that can best predict the sale price of a house.
The goal of this assignment is to test your technical and communication skills. First, you engage in an exploratory data analysis (EDA) which allows you to interrogate the dataset. Correlation analysis between the features is one of the analysis you could do as part of an EDA.
In the case of our house price challenge you might want to take a look at whether the average number of bedrooms varies based on the neighborhood the house is in or try to figure out how the sale price varies based on the area of the garage.
After your analysis, you will conduct a data cleaning exercise and doing any necessary feature engineering that you might need to do.
After which you'll try out a few algorithms to get the job done and this process will be something that's iterative. In these steps it's important to make sure to explain your reasoning behind a choice of model and your method of evaluating it or moving on from it to another model.
The whole point of this task is to take your interviewer through a data science journey because really you're building a story around solving this problem so keep that in mind. And something useful that I found is to hunt down some of these Kaggle challenges and look at the solutions that got the most number of votes because they usually tend to be the ones that have been great at communicating the steps that they've taken and they've also been able to present a fairly reasonable approach to solving the problem. I've found it quite interesting to learn from the best on Kaggle.
So I couldn't recommend it highly enough and in some cases, you might actually get a mix of the online data quiz and also take-home exam I would really encourage you to prepare for all possibilities.
If you're enjoying the video so far consider subscribing. I make weekly videos about demystifying the land of data science.
Onsite Interviews
After you've passed your technical interviews, you'll be invited to interview with a data scientist or lead data scientist at the company.
Before COVID, this would have been in person but these days it's quite normal to expect zoom interviews. In these interviews, there'll be a mix of experience, scenario-based and culture fit behavioral style questions. By this stage, they've generally got a fairly good sense as to your technical abilities.
The first thing they'll want to know is about your work experience. So this is where they might ask you to describe a particular data science problem that you have solved at a previous company or some personal data science project that you have undertaken. In this step, they're really looking to see whether you can describe things in a simple and understandable manner without getting into the weeds because a large part of the role of a data scientist is liaising with different stakeholders.
The next type of questions they'll ask is a scenario-based question they'll want to assess how your problem-solving abilities are regarding specific data science problems in the company that you're interviewing for. Suppose you're interviewing for a social media company with an online platform and they've recently released a new feature which allows users to upload a 10-second gif and they want you to figure out whether this feature has increased user engagement on the platform. Questions like this really require you to fully immerse yourself into the problem space and think practically about an AB style of an experiment. Your answer should be more of a discussion with your interviewer rather than a rigid answer.
Again they want to evaluate your thought process, so don't be afraid to ask clarifying questions. The interviewers want you to succeed. Some of these scenario-based questions might also involve you picking a specific type of model for a type of problem they're looking to learn whether or not you can conduct a successful trade-off between different options while displaying a broad understanding of the various options.
In some of these questions you might feel as though they're poking a hole in your understanding but always be confident. Answer what you can and don't be afraid to admit if you don't know something. This shows them your willingness to learn.
The last thing they want is someone fabricating their understanding of a particular topic because there's really no such thing as fake it till you make it in data science. You'll be working with extremely smart people who'll see straight through you.
Finally, after the scenario-based questions you'll be asked a couple of culture fit questions so this will be things like why you want to work for this particular data science team or at this particular company.
Your answers will be defined by how well you've done your research about the company and the people that work there. So for this step, I always find it useful to go through the company website, look through a couple of the Linkedin profiles of the people that work there and go through any Glassdoor reviews. All of these will give you a good sense as to the culture of the company and also give you a couple of things to talk about and additionally be genuinely curious about your interviewers. Have a look at their experience and prepare a couple of questions to ask them and ideally, these should be questions that are specific about their experience.
For example, it could be What's most rewarding project you've worked on so far at this company? or What's your day-to-day look like? and also ask specific questions about your future role so What does my first three months look like if you get hired? or What type of projects will I be working on? and this will give you a rough understanding as to what you're getting into.
Handy links
https://samanvaykarambhe.com/useful-data-science-resources/
BOOK a 1 on 1 Data Science Career Coaching Session with me - https://calendly.com/samanvaykarambhe/data-science-career-coaching
Check out my other blog posts & YouTube