Agenda
Machine Learning Roundtable: ML in Game Backend
Mathieu Nayrolles (Director of Applied AI/ML, Supercell)
Pass Type:
Academic Pass, Festival Pass, Game Changer Pass, Indie Pass, Start-Up Pass
Hot Topics:
Security
Track:
Programming, Production & Team Leadership
Audience Level:
All
Format:
Roundtable
Vault Recording: TBD
Audience Level: All
Everything player-facing that is not in the moment-to-moment loop. Matchmaking, moderation, player support, anti-cheat, anti-ddos, recommenders. Systems that run behind the game, quietly, for millions of players, and almost never get a public talk.
Part of that silence is deliberate. Nobody wants to hand cheaters a roadmap. A round table is off the record, so it is the right room for it. We will get concrete on what an LLM is actually good for here and what it is too slow or too expensive for. Text moderation is cheap. Anti-cheat wants sub-50 ms and numeric features, not a language model. Player support is where the money moved, with vendors claiming high deflection and studios watching the cost per resolution. And inference cost at scale decides more product questions than anyone admits.
The goal is to compare notes on real deployments. What works, what it costs, and where agents help versus where they just add risk.
Part of that silence is deliberate. Nobody wants to hand cheaters a roadmap. A round table is off the record, so it is the right room for it. We will get concrete on what an LLM is actually good for here and what it is too slow or too expensive for. Text moderation is cheap. Anti-cheat wants sub-50 ms and numeric features, not a language model. Player support is where the money moved, with vendors claiming high deflection and studios watching the cost per resolution. And inference cost at scale decides more product questions than anyone admits.
The goal is to compare notes on real deployments. What works, what it costs, and where agents help versus where they just add risk.
Takeaway
Attendees leave with a practical view of where ML pays off in the game backend and where it does not. This includes matchmaking, moderation, anti-cheat, player support, personalization, and the inference cost that shapes all of them.
Rather than abstract capability, this session focuses on real deployments, real costs per player, and honest limits. Attendees should leave better able to decide what to run at scale, what to keep classical, and where an agent is worth the risk.
Rather than abstract capability, this session focuses on real deployments, real costs per player, and honest limits. Attendees should leave better able to decide what to run at scale, what to keep classical, and where an agent is worth the risk.
Intended Audience
This round table is aimed at ML practitioners working on the systems behind live games. This includes backend and ML engineers, data scientists, trust and safety and anti-cheat engineers, LiveOps and platform teams, and their leads.
You do not need to be an ML expert, but you should work near matchmaking, moderation, support, security, or personalization. If you have shipped or fought with any of these in the past year, this session is for you.
You do not need to be an ML expert, but you should work near matchmaking, moderation, support, security, or personalization. If you have shipped or fought with any of these in the past year, this session is for you.