Back to Open Positions

Physics Engine Developer

Shenzhen

Responsibilities

  1. Own engineering implementation, performance optimization, and stability delivery of core physics engine modules, supporting high-throughput real-time simulation of embodied AI multi-robot and complex scenarios and driving Sim-to-Real adoption
  2. Develop, refactor, and productionize modules including rigid body dynamics, contact/collision detection and response, and geometry processing, ensuring real-time performance and robustness in complex multi-body interaction scenarios
  3. Perform performance analysis, parallelization, and bottleneck optimization for CPU/GPU heterogeneous architectures (C++/CUDA) to support 10k-level concurrency and high-throughput simulation requirements
  4. Build and continuously refine the contact and collision system, solving engineering problems such as non-convex bodies, multi-contact points, and rigid-soft coupling to improve simulation stability and scalability
  5. Collaborate with algorithm, product, and robotics teams to align simulation models with real-world data and participate in Sim-to-Real closed-loop validation and toolchain construction
  6. Follow algorithm implementations and secondary development practices of mainstream physics engines (MuJoCo, Isaac Sim, PhysX, etc.) and quickly productionize mature methods

Requirements

  1. Master's degree or above in Computer Science, Mechanics, Physics, or a related field, with 2+ years of work experience in physics simulation/engines (requirements may be relaxed for outstanding candidates)
  2. Proficient in C++ and Python, with solid CUDA programming and GPU parallel optimization experience, and able to independently convert algorithm prototypes into high-performance engineering code
  3. Deep understanding of rigid body dynamics, computational geometry, and numerical methods; hands-on development or deep secondary development experience with complex-shape contact/collision algorithms, or engineering experience in fluid/deformable body simulation
  4. At least deep usage or secondary development experience with one mainstream physics engine (MuJoCo, Isaac Sim, PyBullet, PhysX, etc.), with a clear understanding of the underlying implementation and the ability to modify source code
  5. Strong mathematical foundation (linear algebra, numerical analysis, computational geometry) and ability to independently locate and resolve numerical stability and performance issues
  6. Strong problem decomposition and engineering delivery skills, with good communication and collaboration awareness and the ability to work effectively with cross-background teams

Bonus Points

  • Practical experience in large-scale parallel (100k to 1M rigid bodies/particles) performance optimization
  • Familiarity with differentiable physics or RL simulation training pipelines, or Sim-to-Real engineering experience
  • Some understanding of robotics, including kinematics/dynamics modeling and contact dynamics
  • Contributions to open-source physics engine core modules or experience implementing complete submodules from scratch