About the job
Join Our Innovative Robotics Team
At OpenAI, our Robotics team is dedicated to pioneering general-purpose robotics and striving towards artificial general intelligence (AGI) within dynamic, real-world environments. By integrating advanced hardware and software, we explore a diverse array of robotic configurations, aiming to harmoniously combine high-level AI capabilities with the physical constraints of real-world systems to enhance everyday lives.
Your Role as a Simulation Environments Engineer
We are seeking a talented Simulation Environments Engineer to develop the tools and infrastructure that facilitate the creation of comprehensive, realistic virtual environments for robotics research and evaluation. This position emphasizes building systems that empower researchers and engineers to define, visualize, generate, and validate task environments efficiently and at scale. You will design processes for importing and validating third-party content, develop procedural and randomized scenario generators, and produce user-friendly tools to streamline environment creation, making it swift, repeatable, and testable. This role sits at the convergence of game-engine expertise, asset engineering, and extensive simulation infrastructure.
Location and Work Schedule
This position is based in San Francisco, CA, requiring in-person collaboration three days a week.
Key Responsibilities
Develop interactive and programmatic tools for researchers to describe, preview, and validate scenes and tasks, enabling quick and consistent scenario authoring.
Establish content pipelines to curate, convert, optimize, and quality-check visual and collision assets from both external collections and internal resources, ensuring consistent behavior across engines and tasks.
Create robust importers and adapters to integrate environments and setups from platforms such as Isaac, Unity, Unreal, and Omniverse into our simulation pipelines, maintaining fidelity and performance.
Develop frameworks for procedural generation and controlled randomization (covering visual, physical, and kinematic aspects) to provide models with systematic, measurable variations in conditions.
Define and uphold quality standards for environments (including visual fidelity, collision accuracy, and physical realism) and implement validation tools to ensure environments meet established realism and coverage objectives.
Link environment tools to CI/CD processes, presubmit checks, large-scale simulation farms, and model evaluation pipelines to facilitate automatic testing and large-scale execution of environments.
