World Model / Action Policy Researcher at Modal

New York, New York, United States

Apply Now

$350,000 – $450,000Compensation

Senior (5 to 8 years)Experience Level

Full TimeJob Type

UnknownVisa

Gaming, Artificial Intelligence, TechnologyIndustries

Requirements

5+ years of experience in deep learning research or reinforcement learning, with a focus on embodied agents or simulation environments
Strong foundation in representation learning and generative modeling, particularly using architectures such as diffusion models, VAEs, and transformers applied to video
Experience with world models and predictive control — understanding how to train models that simulate dynamics and plan actions in learned environments
Proficiency in reinforcement learning (RL, model-based RL, or imitation learning) and the ability to design and evaluate policy networks
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 or simulations
Publications or open-source contributions in areas like world modeling, simulation learning, or agent policies (strong plus)
In-person work in NYC, 5 days in the office
Ownership & scientific rigor: Seeing ideas through from concept to proof to deployment, writing clean, reproducible code, and maintaining high experimental validity
Performance and scaling mindset: Understanding compute efficiency, distributed training, and data bottlenecks
Curiosity-driven and result-oriented: Excited by open-ended problems, defining measurable goals, and shipping impactful systems
Gaming & simulation passion: Interest in interactive environments, physics-based simulations, or gaming AI; experience with Unity, Unreal Engine, or custom simulators (plus)

Responsibilities

Designing, training, and evaluating world models and action policies that operate within games and based on gaming data
Experimenting rapidly and iterating on architectures
Collaborating closely with product and engineering teams to bring research ideas to production
Working with a team to contribute to intelligent agents and simulation systems that understand, predict, and act in complex 3D environments

Skills

Python

PyTorch

deep learning

reinforcement learning

world models

diffusion models

VAEs

transformers

generative modeling

representation learning

embodied agents

simulation environments

model-based RL

imitation learning

policy networks

Modal

Employee training and skill development platform

About Modal

Modal Learning focuses on improving employee performance through skill development for businesses. Their main product, the Modal Mastery Platform, uses active learning techniques, including live cohort sessions, labs, and one-on-one coaching, to help employees engage with the material effectively. Unlike competitors, Modal Learning offers a subscription model that provides structured eight-week training programs, aligning skill development with organizational goals. The company's goal is to empower employees and help organizations retain talent by providing clear career development paths.

San Francisco, CaliforniaHeadquarters

2021Year Founded

$30.8MTotal Funding

EARLY_VCCompany Stage

Consulting, EducationIndustries

51-200Employees