Data Science — How to prepare for the interview
To say that Wildlife has been very successful over the course of its 9-year existence is uncontroversial. Our more than 70 games have been downloaded two billion times in almost every country, making us one of the top 10 mobile game developers and publishers in the world. This is a testament to the team’s ability to consistently produce experiences that can engage huge numbers of players, in a way that also makes for a healthy, profitable business.
This is still the beginning of a much more ambitious journey. We aim to develop games that will impact an entire generation of users, and products that will fundamentally change the ways in which technology and entertainment combine. To accomplish those unabashedly big goals, we need the very best people — including, but not only, in data science. A great interview process is the first step in that direction.
Data Science at Wildlife
Every aspect of the way we develop, iterate on, and market our products is informed by data. We want our games to become massive hits — which means we must leverage hundreds of terabytes of data to make them more engaging, fun, profitable, and able to reach their target audience effectively.
There are probably more use cases for data science at the company than I could count.
We have scientists working on improving the matchmaking system of Tennis Clash, calibrating the player-versus-player (PvP) mode of Sniper 3D to integrate it more seamlessly into the game’s core experience, applying supervised learning models to predict the lifetime value of different user segments, designing better algorithms to do real-time bidding, etc.
What all of those projects have in common is that they’re challenging from a technical standpoint and produce a large impact for the company.
If you would like to get a richer understanding of the role that data scientists play at Wildlife, be sure to check my previous article on Why Great Mobile Gaming Companies Have Data Scientists.
What we look for in data scientists
We’re big believers in the idea of hiring full-stack data scientists — generalists over specialists. This deserves a post on its own, and, fortunately, Eric Colson did a wonderful job explaining why Stitch Fix, a fashion company known to have one of the very best data organizations in the world, embraces that philosophy.
The basic idea is that our data scientists are typically focused on specific products over which they have a lot of autonomy and control — but also very broad responsibilities. This makes the learning process easier, reduces coordination costs and allows us to move fast. On the other hand, it requires two things:
- That data scientists are able to perform a wide variety of activities, from translating business needs into mathematical form all the way through deploying machine learning models at scale.
- That they can rely on a set of tools and processes that are designed to make their work more efficient by abstracting away some of the engineering complexity (related to infrastructure management, CI/CD, etc.).
It should be clear by now that, by saying we look for full-stack data scientists, we do not mean we’re after people who must also be professional data engineers and professional machine learning engineers. In fact, collaboration with those two functions — which, I repeat, are distinct from our own — is essential to our success.
To be consistent with that philosophy, the technical expertise we require from candidates during our interview process revolves around the foundations of data science — statistics, machine learning, programming — rather than specific details of algorithms and technologies.
Don’t know how TensorFlow Lite allows you to perform inference in deep learning models on mobile devices? No problem — you’ll never get asked that sort of question. Don’t know the difference between two fundamental ways to describe the performance of a classification model? Then you might be in trouble.
In addition to technical knowledge, we value people who adhere to our culture — people who are intellectually curious, confident but aware of the limitations of their knowledge, driven by impact, ambitious yet collaborative, and able to communicate clearly.
Our interview process step by step
Our interview process is based on three operating principles:
- It contains a series of structured, clearly defined steps, because we believe this makes it more reliable as a predictor of future performance and more inclusive in the sense of not leaving much room for unconscious bias.
- It has a high bar for both talent and culture fit. Wildlife aims to be the place where every outstanding contributor can achieve their personal best, and we know that one of the things that make talented, driven people flourish is to be surrounded by other talented, driven people.
- It is fast because everyone hates waiting for several weeks for feedback. Indeed, we believe speed is key to providing candidates with a superior experience.
Small details can vary by location, hiring manager, and level of seniority, but overall we invest a lot of effort into making sure our hiring standards are consistent, given how fundamental this is to build a high-performance distributed organization.
In general, first-round interviews are conducted by a potential future teammate. They tend to have a very broad scope, encompassing your experiences and projects, personal interests, what you cherish vs. dislike in the workplace, etc. It can also include a couple of open-ended questions designed to assess your conceptual thinking, as well as some technical questions.
Second-round interviews are purely technical and revolve around statistics, machine learning, and engineering tools applied to data science. They might include a practical case for you to solve. You may feel you don’t know the answers to a number of them — that’s normal. Getting some questions wrong will not be enough to disqualify you.
Final-round interviews are always done by a manager — or senior manager, or director, or even the CTO. They’re focused on culture fit. We aim to understand if Wildlife has the sort of environment where you would both be successful and elevate those around you. When you value culture as much as we do, you have to actively select for it.
If all goes well, you get an offer and, sure enough, you accept it, join the company, and everyone lives happily ever after.
How to prepare
Interviews are supposed to help us get to know you as you already are, but it does no harm to prepare a bit for the technical part. We’re all about strong foundations, so the learning resources we’d recommend have that explicit focus.
When it comes to books, An Introduction to Statistical Learning is a great and accessible volume on machine learning, and Statistical Rethinking is a personal favorite for many of us, myself included. Pythonistas typically write more practical books, among which Data Science from Scratch and Python Data Science Handbook are both very good choices, albeit very different from each other. There are also a number of good online courses, among which Coursera’s Machine Learning by Andrew Ng is a classic.
Regarding the interview experience itself, we know that this is probably not the most relaxing activity for most people. With that in mind, take the time to do whatever may help you feel more comfortable before each round of conversations. Keep in mind that this is what it’s about: having meaningful, two-way conversations designed to help us assess whether you match what we’re looking for, and also to help you think about whether Wildlife is the sort of company you’d like to work at. Ask as many questions as you want — we’re always happy to answer them.
I hope this post has provided you with a good overview of what we expect from data scientists at Wildlife and what our interview process looks like. If it sparks your interest, be sure to check out our job board!