Position Overview
We are seeking a Physics Simulation Scientist to lead advancements in the simulation and physics-solving backbone behind Skild’s robot foundation model training. You will collaborate with external experts in GPU-accelerated physics engines and work with our internal robotics and learning teams to build a next-generation, open-source simulation stack for robotics sim-to-real.
You’ll partner closely with engineers scaling simulation scene generation and with ML researchers pushing the limits of sim-to-real transfer. The ideal candidate brings deep physics-simulation expertise plus hands-on experience implementing and optimizing algorithms on modern GPUs.
Responsibilities
- Improve and develop new physics solvers and modeling methods for high-DoF, contact-rich robotics tasks.
- Design and implement GPU-accelerated solvers with a focus on throughput, stability, and scalability.
- Profile and optimize simulation performance on modern GPU hardware and distributed clusters.
- Work with external collaborators and the open-source community to advance simulation for robotics.
- Collaborate with scene-generation engineers to scale robotic experience across diverse real-world environments.
- Partner with ML researchers to improve sim-to-real transfer through better physical modeling, calibration, and training regimes.
- Contribute to the long-term technical direction of Skild’s physical modeling and sim-to-real strategy.
Preferred Qualifications
- MS or PhD in Physics, Robotics, Computer Science, Applied Math, Engineering, or a related field, or equivalent hands-on experience.
- Strong track record working on physics engines or high-fidelity simulators for articulated rigid bodies; experience with deformables, fluids, or differentiable simulation is a plus.
- Deep understanding of dynamics, contact modeling, constraint-based methods, and integrators, including accuracy–speed tradeoffs.
- Expertise in CUDA and GPU programming with proven ability to optimize for scale.
- Proficiency in C++ and Python, and experience building reliable systems used by other technical teams.
- Familiarity with how modern ML/RL pipelines consume simulation (vectorized environments, domain randomization, large-scale rollouts).
- Experience with real robot platforms and strong intuition for where simulation diverges from reality.
- Publications, open-source contributions, or shipped systems in simulation, robotics, graphics, or numerical computing are a strong plus.
