Staff Research Scientist, Foundation Models - Tech Research

Riot Games•Los Angeles, CA

About The Position

Research Scientists at Riot combine deep technical expertise across machine learning, artificial intelligence (AI), experimental design, and systems engineering to invent new ways players experience games. As a Staff Research Scientist, you push the boundaries of what’s possible and raise the bar for research quality and its translation into extraordinary gameplay. The Technology Research team performs applied research to accelerate the pace of building amazing player experiences at Riot Games. As a Staff Research Scientist, you will research and develop world models—learned representations that understand, predict, and simulate game environments. Rather than building general-purpose large language models, you will focus on models that capture the structure and dynamics of Riot's games: how state evolves, how players behave, and how design changes ripple through complex systems. Your work will power scalable simulation for agent training, automated game balancing, synthetic data generation, and new forms of language-conditioned game interaction. You will stay deeply hands-on, turning advances in predictive simulation, latent state representations, and environment modeling into concrete prototypes that game teams can evaluate and build on. You will work closely with game teams to ground your research in real gameplay needs and partner across Technology Research to connect world models with game AI and NLP efforts.

Requirements

Ph.D. in Machine Learning, AI, Statistics, Math, or related Computer Science/Quantitative field, or equivalent experience (i.e., M.Sc. with 3+ years of relevant experience, or B.Sc. with over 6+ years of relevant experience, etc.)
2+ years experience in deep learning or algorithmic foundations of optimization
Experience with generative models, representation learning, model-based reinforcement learning, or learned simulation
Proficiency in C, C++, or Python
Expertise with TensorFlow or PyTorch

Nice To Haves

Experience building or training world models, dynamics models, or latent-space simulators (e.g., Dreamer, MuZero, or similar architectures)
Familiarity with game engines (e.g., Godot, Unity, or Unreal) for prototyping or data collection
Distributed computing systems design or large-scale model training infrastructure
Experience with GCP or AWS
Demonstrated contributions to public research (e.g., publications, talks, open-source projects)

Responsibilities

Propose, implement, and evaluate world models that learn latent representations of game state, player behavior, and environment dynamics
Build predictive simulations that enable agent training, game balance testing, and rapid design iteration without full game server overhead
Develop methods for synthetic data generation and auto-balancing powered by learned environment models
Explore language-conditioned approaches that combine large language models with state estimators to enable new forms of game interaction
Work closely with game teams to understand gameplay needs and design constraints, ensuring research is grounded in real player experiences
Partner with game AI and NLP researchers to integrate world models into broader AI systems across Technology Research
Stay current with academic and industry advances in world models, predictive simulation, representation learning, and environment modeling (e.g., MuZero, Gato, Dreamer)
Communicate progress and learnings clearly to stakeholders, helping teams understand both what works and what doesn’t