Breaking into Data Science
You may have heard the classic catch-22 when it comes to job hunting: “You need experience to get experience”. And sadly, I considered it to be true for a while.
While I had a couple of gigs in between, it took almost a year and a half for me to find my first real job as a data scientist. It was a tiring and depressing process, and I absolutely questioned by self-worth during most of it.
Here, I’m going to talk about my experience in the data science interview process and hopefully provide some insight into how to break through.
Why is it so hard?
1: Data science is a young field
Data science as a career path has only been around since 2008. And because it’s still fairly new to a lot of industries, many companies are not adequately prepared to train and mentor their own data scientists. As a result, not only are junior-level opportunities in data science harder to come by, but that fosters more competition for the junior-level opportunities that DO exist.
2: Inconsistent job requirements
Data science roles and responsibilities exist on a spectrum. On one extreme is business analysts, that spend much of their time building dashboards and communicating with stakeholders. On the other extreme is machine learning engineers, configuring deep learning models in Tensorflow for the likes of speech recognition or self-driving cars. The skills required for each are very different, but both can be considered as “data scientist” depending on the company.
This article discusses some of the unrealistic expectations that show up in data science job listings as a result of the wide variety of skillsets.
3: Data science emphasizes product sense and business value.
Data science is not just about knowing the ins-and-outs of machine learning algorithms to solve a given prediction problem; it’s also about determining which problems drive the most value for the business. These are skills that are hard to learn and demonstrate without having worked in a business environment.
The interview process
There’s typically 3 steps to a data science interview process before getting to the offer stage:
1: Phone screen with a recruiter; 30 minutes
The purpose of this interview is to ensure that your background, interests, salary expectations, etc. are consistent with the company’s objectives. It is usually done with a recruiter who is not likely vet your technical knowledge too heavily. Instead, they are more interested in assessing communication as well as “culture fit.” In other words, would you enjoy working with them, and vice versa?
This is a great opportunity to ask questions about the company and their culture, structure, values, etc.
2: Technical interview; 1 hour
The goal of this interview is not only meant to assess your coding skills, but also communication skills and your ability to think programmatically. This is typically done 1:1 with an engineer/data scientist at the company, and is usually conducted in Python or SQL. Ideally, you’ll have time at the end of the interview to ask any remaining questions
3: Onsite; 3+ hours
This usually consists of 4+ interviews covering a wide array of topics. All of your interviewers may be data scientists, or you may have product managers and software engineers interviewing you as well. This typically depends on the size/structure of the company, as well as the nature of the specific interview.
I see 3 different types of interviews at this stage, and each one has enough depth to merit its own article:
Behavioral: You’ll hear questions about soft skills. For instance, how you work in a team environment, handle conflict, and communicate.
Technical: This will be similar in concept to the technical interview you saw in the previous stage, but is usually longer and involves more follow-up questions. While you will most likely see a coding question here, that will not encompass the full interview. You may also hear questions about concepts related to machine learning, statistics, or product sense.
Past Project: Several times in data science interviews, I’ve been asked to prepare a presentation on a data science project I’ve worked on. I honestly like these interviews, because this is one of the few places in the interview process where you are in full control; where you can showcase your strongest skills.
My current employer posted this guide as well, which is fairly consistent with what I’ve noted here.
Tips and Tricks
It’s significantly harder to find a job or even get interviews as a data scientist if you haven’t had experience in the space before.
Here are some ways to bridge that gap:
1: Personal data science projects
These are a great way to learn and demonstrate skills with the endless list of data science tools, as well as fill up a portfolio in job applications. Kaggle provides a lot of free datasets which can serve as opportunities to practice exploratory data analysis (EDA) and various supervised machine learning models. And like I mentioned before, these come up fairly often in data science interviews.
2: Freelance
Freelancing on platforms such as Upwork not only provides a means of paying the bills, but also a way of lowering the entry barrier to officially calling yourself a data scientist. It also serves as an opportunity to essentially run your own business, which utilizes skills such as personal discipline, time management, and stakeholder communication; all of which are important soft skills for any data scientist.
Parting Thoughts
I often hear people compare job searching to dating, and I’m someone who has trouble with both. Ultimately, both are about assessing fit, testing each other, and determining whether or not you support each other’s various needs.
In an interview, it is not just the company interviewing you, but you are also interviewing the company. And if an interview doesn’t pan out, then it’s not always the case that you screwed up. It’s also possible that you might not have been happy in that company/role anyways.
