About the job
About Our Team
Join our innovative Robotics team at OpenAI, where we are dedicated to pioneering general-purpose robotics and advancing towards AGI-level intelligence in dynamic, real-world environments. By working across the entire model stack, we seamlessly integrate state-of-the-art hardware and software to explore a wide variety of robotic form factors. Our mission is to harmoniously combine advanced AI capabilities with the constraints of physical systems, ultimately enhancing the lives of individuals worldwide.
About the Role
We are on the lookout for a passionate Simulation Realism Engineer to lead our initiative in achieving quantitative realism in simulation for robotics research and product development. You will be responsible for devising strategies, tools, and operational practices that bridge the sim-to-real gaps across physics, sensors, and rendering. Collaborating cross-functionally with research, software, hardware, and operations teams, you will identify key realism gaps, establish measurable realism metrics, and deliver impactful engineering solutions — from fine-tuning engine parameters and setting asset standards to integrating third-party engines and managing simulations at scale. This hands-on engineering role combines scientific precision (measurement & validation) with practical systems work (tooling, cloud ops, and CI/HIL pipelines).
This position is based in San Francisco, CA, with a requirement of 4 days in the office per week. Relocation assistance is available for new hires.
Your Responsibilities:
Establish and operationalize realism metrics & protocols. Design experiments and automated tests to identify specific areas of non-realism, quantify gaps, and monitor regressions over time.
Optimize engine parameters (contacts, friction, mass/density, solver settings) and object models to ensure simulated dynamics align with real-world measurements.
Assess, integrate, and where necessary, extend third-party physics, rendering, and sensor simulation engines (e.g., Isaac, PhysX, MuJoCo). Lead vendor proof of concepts and benchmark features to inform roadmaps.
Develop standards and guidelines for authoring and validating asset physical/visual properties; implement semi- or fully-automated pipelines to tune and optimize assets for target engines.
Resolve technical challenges related to running engines in the cloud (OS, drivers, GPU), parallelize simulations (batch multiple runs per engine instance), and fortify real-world-facing pipelines to ensure reliable simulation at scale.
Lead validation campaigns (teleoperation, HI...) to verify simulation accuracy and performance.
