Member of Technical Staff – Research

Roboticsconnected; New York (NY); On-site; USD 180,000 - 250,000; Full time

Job summary

General Intuition is seeking researchers to train action policies and world models on large, labeled gameplay data. You will build policies that operate from pixels and condition on language, driving from concept to proof to deployment.

Applicants should have extensive experience in deep learning, especially imitation learning, reinforcement learning, and representation learning. The role stresses production-minded research and scalable, efficient training systems.

Qualifications

• Experience with imitation learning, RL and world modeling at scale.
• Publish or contribute to open-source in agent policies, imitation learning, or world modeling.
• Experience training policies from pixels on large, messy real data.

Responsibilities

• Train action policies and world models from pixel data.
• Develop scalable training pipelines and ensure reproducibility.
• Bridge research into production, focusing on compute efficiency and data bottlenecks.

Skills

Imitation learning

Reinforcement learning

World modeling

Diffusion models

Transformers

PyTorch

Python

Tools

Python

PyTorch

Job description

ABOUT THE COMPANY

General Intuition is the frontier lab for acting in space and time. We build large action models and world models that can perceive, predict, and act across virtual and physical environments. General Intuition builds on the strength of Medal, the world's largest and fastest-growing platform for gaming clips, where millions of gamers capture, share, and discover new games every year. We've raised over $650M from Khosla, GC, Valor, and Point72 since October 2025, and recently closed our latest round at a $6.2B valuation.

THE ROLE

We train action policies and world models on the largest corpus of action-labeled gaming data in the world. You'll train policies that act from pixels on our action-labeled gameplay data, including policies conditioned on language, and see ideas through from concept to proof to deployment. You'll write clean, reproducible code, keep a high bar for experimental validity, and care about how research translates into production systems: compute efficiency, distributed training and data bottlenecks. We're looking for people who are excited by open-ended problems but know how to define measurable goals and ship impactful systems.

WHAT WE'RE LOOKING FOR

• Proficiency in imitation learning, reinforcement learning and world modeling at scale, and the ability to design and evaluate policy networks.
• 5+ years of experience in deep learning or reinforcement learning research, with a focus on action policies or embodied agents.
• Strong foundation in representation learning and generative modeling, particularly diffusion or flow-matching models and autoregressive transformers.
• Publications or open-source contributions in areas like agent policies, imitation learning, or world modeling.
• Experience training policies from pixels on large, messy real data, not only in simplified settings.
• Programming fluency in Python and deep learning frameworks such as PyTorch.
• Strong experimental skills - comfort with large-scale training, evaluation pipelines, and managing complex datasets.
• First-principles fundamentals: compute cost, what a loss value tells you, fair comparisons at equal compute.
• Gaming passion: Interest in games or gaming AI.