Career Tracks in ML in the Games Industry

Machine Learning (ML) in games isn’t just hype anymore — it’s a real chance to build a career in a fast-moving field. Whether it’s marketing, recommendation systems, player behavior prediction, or R&D, ML already opens up plenty of paths for engineers and researchers.

We sat down with Dmitrii Sigalov, ML Product Owner at SayGames, to talk about the different career tracks in ML, what makes this role special in game development, and where this journey can take you.

— Dmitrii, how did you get into ML?

— I started experimenting with ML about three years ago, right when the big hype around ChatGPT kicked off. Before that, I’d spent a long time working in game development as a programmer and tech lead. Initially, I wrote some very simple models “from scratch” to understand how they worked, and then I progressed to using proper frameworks. What kept me in ML was the chance to explore new things and immediately see the practical impact of what I was building.

— Where is ML used in games nowadays?

— The most common use of ML in games right now is in marketing — things like LTV prediction, cohort analysis, and revenue forecasting. Almost every company with a marketing team needs long-term cohort predictions, and while some rely on traditional algorithms, others are already turning to ML for more accuracy.

But there’s another area that feels far less explored: applying ML inside the game itself. Think recommendation systems, predicting player behavior, and adapting gameplay in real time based on those predictions.

Some studios are already experimenting with this, but I believe the biggest opportunities are still ahead. Personalized in-game experiences, smart recommendations, dynamic gameplay — that’s where I see the true future of ML in games.

— People often say ML models are harder to interpret than traditional algorithms. Is that true?

— In many ways, yes. ML is often seen as a “black box.” For someone who isn’t deeply into ML or the math behind it, it can feel like pure magic: you put something in, and results come out.

But for someone with the right technical background, it’s less mysterious. At its core, it’s really just about vectors and how they relate to each other. A model provides the result that’s closest to reality, which is why the prediction is effective.

Unlike algorithmic methods — where you can usually keep the key parameters in your head — ML involves far too many factors influencing the outcome. That’s why it’s much harder to interpret. Still, it’s workable. As long as you can test against a baseline and see whether the model performs better or worse, you can move forward. The real challenge is that when something goes wrong, it isn’t always clear whether the issue lies in the model, the data, or even the expectations.

— What career tracks do you see in ML? Where do people usually start, and where can they grow?

— In the games industry, a few main paths are already visible:

• The classic engineering track: Engineer → Senior → Lead.
• The R&D track: Research → Applied Research → R&D Lead.
• The ML Ops / Platform track: from data and infrastructure work to Architect.
• Management and product: ML Product Manager or other leadership roles.

However, the truth is that there’s no single “right” path. People enter ML from very different backgrounds. Good entry points are roles that already work with large datasets — data analysts or data engineers. In ML, you can’t avoid data: you need to know how to prepare it, work with it, and analyze the results.

That’s why every data scientist has to be a bit of an analyst and a bit of an engineer. You also need at least basic programming skills, typically in Python, and proficiency in SQL for data engineering.

As for what comes after ML — most specialists grow into team leads or heads of directions. C-level roles coming directly from ML are still rare, simply because even in large companies ML teams remain relatively small. However, the field is still young, and I believe we’ll see even bigger leadership opportunities in the future.

At SayGames, our ML team is growing, and the diverse backgrounds of our colleagues demonstrate the numerous entry points into this field. One teammate started out as a data scientist, another joined from a role focused on long-term LTV predictions, and another transitioned from programming thanks to a strong academic foundation in data science. All of them found their own path into ML — and that’s exactly the point: there isn’t just one way to build a career here.

— What is your team working on right now?

— Our main focus is on building an auction-based ad network, where predictions are made in real time. It’s a high-load system, and we’re looking for specialists with experience in ad networks or recommendation systems.

What makes ML different from many other areas of development is that R&D is almost always part of the process. Every new challenge usually starts with research: first you need to understand what data is available and what can be done with it, and only then do potential solutions begin to take shape. That approach not only drives innovation but also helps specialists grow, without getting stuck repeating the same type of tasks.

In our team, approximately 10% of the time is dedicated to in-game predictions — tasks such as forecasting churn or identifying potential payers, and then designing mechanics that utilize these predictions to enhance the player experience. It’s an exciting direction, and we’re already seeing promising results that we’re eager to take further.

— Why is now a good time to join your team?

— Right now, the team is still young, and many processes aren’t set in stone yet. That means there’s a real chance to shape how things are built, influence decisions, and help create something new. A lot has already been done, but there’s still so much ahead — and it’s always more exciting to build from the ground up than to step into a rigid structure where nothing can be changed. With us, it’s not a set of rails you have to follow — it’s more like a trail that’s only just being made, and you can help decide where it leads.

— And what do you look for in candidates? What makes a good match?

— Of course, it would be great if someone brings the right expertise, but even more important is the mindset. We’re looking for curious people, who want to experiment, explore, and build in an environment where not everything is defined yet. A willingness to work in the unknown and figure things out from scratch is key.

— What advice would you give to newcomers and experienced specialists?

— For newcomers, it’s tough right now in the industry. Almost everything a beginner can do, GPT can already do. That’s why the best approach is usually through adjacent fields — analytics or data work that sits close to prediction tasks, with a gradual transition into ML itself.

For mid-level and senior specialists, the difference is mostly about ownership. A mid-level needs direction on what to do, while a senior can find a problem and figure out how to solve it. My advice for both: develop a stronger business perspective. Don’t just close tickets — try to see the bigger picture, understand the underlying problem, and solve it in a way that creates real value.

— And what inspires you personally about the future of ML in games?

— I’d love to see ML used to create a more personalized in-game experience. Currently, predictions are primarily used in marketing analytics. But the idea of a game that adapts to each player — offering what they need, keeping them engaged, and making their journey unique — that’s where I see the deeper future of ML in gaming.

At SayGames, we believe ML will shape the future of games, and we’re excited to be building that future now. The role of ML in gaming is only beginning to unfold, and the most exciting breakthroughs are still ahead. If you’re passionate about ML and games, we’d love to hear from you.