Research Scientist V

Owns a physics-based motion reward used to post-train generative video models via reinforcement learning. Three work streams: improving the dynamic and contact terms of the reward signal; making the reward cheap enough to run inside the loop at scale rather than as an offline filter; and extending it from single-body plausibility to human-object and human-scene interaction using 3D mesh representations.

• Rigid-body/contact simulation (Mu Jo Co , Isaac, Bullet or equivalent)
• 3D human pose and motion representations (SMPL-family)
• Python/PyTorch with profiling and optimization of compute-heavy pipelines
• Reward design and post-training of large generative models
• Numerical methods and solver stability

MS/PhD in robotics, graphics, mechanical engineering or CS, or equivalent industry experience

• Publications in physics-grounded generative modeling or differentiable simulation
• Generative video/diffusion experience
• Scaling simulation workloads
• Validating simulated physical quantities against human-judged ground truth

Pay range is $79 - $84 per hour with full benefits available.

• paid time off
• medical/dental/vision/life insurance
• 401K
• parental leave
• and more

Our compensation reflects the cost of labor across several US geographic markets. Pay is based on several factors including market location and may vary depending on job-related knowledge, skills, and experience.

At Crystal Equation, we empower people and advance technology initiatives by building trust. Your recruiter will prep you for the interview, obtain feedback, guide you through any necessary paperwork and provide everything you need for a successful start. We will serve to empower you along the way and provide the path for your professional journey. For more information regarding our Privacy Policy, please visit