Research Scientist

Vertisystem (A MOURI Tech Company); United States; On-site; USD 150,000 - 210,000; Full time

Job summary

Vertisystem (A MOURI Tech Company) is seeking an experienced Research Scientist to advance physics-based motion modeling and generative video research. You will develop and optimize physics-based motion reward systems used to post-train generative video models through reinforcement learning, working at the intersection of physics simulation, 3D human motion, RL, and AI-driven video generation.

The role demands 5+ years of hands-on work, a MS or PhD preferred, and strong Python/PyTorch skills;

Qualifications

• 5+ years of relevant research or industry experience.
• MS or PhD in CS/Robotics/CG/ME or related field (or equivalent experience).
• Strong experience with rigid-body and contact simulation tools (MuJoCo, NVIDIA Isaac, Bullet).
• Experience with SMPL-family human body representations and 3D pose modeling is desirable.
• Strong programming skills in Python and PyTorch; experience in reinforcement learning.

Responsibilities

• Develop physics-based motion reward systems for post-training generative video models.
• Integrate reward functions into RL training pipelines at scale.
• Extend motion modeling to human-object and human-scene interactions.
• Work with 3D mesh representations and human motion models to improve realism.
• Profile and optimize compute-intensive simulation and training pipelines.
• Apply reinforcement learning techniques to post-train large generative models.
• Contribute to research in video generation and physics-grounded AI systems.
• Evaluate simulated behavior against ground-truth benchmarks.

Skills

Python

PyTorch

Reinforcement learning

Rigid-body simulation

Physics-based modeling

Education

MS or PhD in Computer Science, Robotics, Computer Graphics, Mechanical Engineering, or related field

Tools

MuJoCo

NVIDIA Isaac

Bullet

Job description

Duration: 12 Months Contract (possible extensions or conversion to FTE)

Experience: 5+ Years

Education: MS or PhD preferred

This is a W2 role, please do not apply if you want to work on C2C or 1099.

About the Role

We are seeking an experienced Research Scientist to advance physics-based motion modeling and generative video research. In this role, you will develop and optimize physics-based motion reward systems used to post-train generative video models through reinforcement learning.

You will work at the intersection of physics simulation, 3D human motion, reinforcement learning, and generative AI to improve the physical realism of AI-generated videos.

Key Responsibilities

• Develop and improve physics-based motion reward systems, including dynamics and contact modeling.
• Integrate reward functions directly into reinforcement learning training pipelines at scale.
• Extend motion modeling capabilities to human-object and human-scene interactions.
• Work with 3D mesh representations and human motion models to improve physical plausibility.
• Profile and optimize compute-intensive simulation and training pipelines.
• Apply reinforcement learning techniques to post-train large generative models.
• Contribute to research in advanced video generation and physics-grounded AI systems.
• Evaluate simulated physical behavior against human-judged ground truth and relevant benchmarks.

Required Qualifications

• 5+ years of relevant research or industry experience.
• MS or PhD in Computer Science, Robotics, Computer Graphics, Mechanical Engineering, or a related field; equivalent industry experience will also be considered.
• Strong experience with rigid-body and contact simulation using tools such as MuJoCo, NVIDIA Isaac, Bullet, or equivalent frameworks.
• Experience with SMPL-family human body representations, 3D human pose modeling, or human motion representations.
• Strong programming skills in Python and PyTorch.
• Experience with reinforcement learning, reward function design, or generative model post-training.
• Experience profiling and optimizing compute-intensive workloads and ML pipelines.
• Understanding of numerical methods, solver stability, and physics-based modeling.

Preferred Qualifications

• PhD in a relevant technical field.
• Research experience in physics-grounded generative modeling or differentiable simulation.
• Experience with generative video models, diffusion models, or video synthesis.
• Experience scaling simulation workloads and distributed computation.
• Publications in relevant research areas.
• Experience validating simulated physical quantities against real-world or human-evaluated ground truth.

Ideal Candidate Profile

The ideal candidate combines expertise in physics simulation, reinforcement learning, 3D human motion, and generative AI. Candidates with backgrounds in robotics simulation, computer graphics, embodied AI, human motion generation, or physics-based machine learning are encouraged to apply.