Agenda

From classic ML to more recent generative approaches, we keep trying to push ML further into the runtime of our games and create delightful experiences for the players. Sometimes it works. Sometimes it barely works. Sometimes it really should not have worked at all.

This round table is about sharing what we actually managed to ship at runtime last year. What made it on screen. What survived real frame budgets, memory constraints, and certification. And what we had to kill along the way. Expect concrete examples, uncomfortable tradeoffs, small wins, and a few embarrassing failures.

This is not a theory discussion. It is a conversation about real systems running in real games.

Takeaway

Attendees leave with a clearer picture of what it really takes to ship modern ML at runtime in games. This includes concrete lessons on performance constraints, system design tradeoffs, debugging strategies, and when ML genuinely beats classical approaches and when it does not.

Rather than best practices or recipes, attendees gain practical intuition from peers about what worked, what failed, and why. The goal is to help you make better decisions the next time you consider putting ML directly in a game loop.

Intended Audience

This round table is intended for ML practitioners and game developers who are already experimenting with or seriously considering runtime ML. This includes gameplay programmers, AI engineers, technical designers, rendering and animation engineers, and tech leads.

An intermediate understanding of ML is helpful, but curiosity and hands-on experience matter more than formal background. If you have tried to ship something ML-driven at runtime or are about to, this session is for you.