Genesis-World: Core Physics Engineer

Genesis•San Carlos, CA

13h

About The Position

Genesis-World is an open-source, general-purpose simulation platform for physical AI from Genesis AI. It features a unified multi-physics engine supporting rigid bodies, FEM, MPM, particles, cloth, and fluids, allowing complex interactions within a single simulation. The platform includes Nyx, a high-performance renderer for robotics, and advanced sensor simulations. Genesis-World is under continuous development, with a focus on creating a comprehensive and fast Incremental Potential Contact (IPC) solver for deformable body dynamics. The engine powers real business applications and is designed to run anywhere, with kernels written once and compiled for various backends (CUDA, AMD ROCm, Apple Metal, Vulkan, x86, and ARM64) by their in-house JIT compiler, Quadrants. This enables massively batched GPU simulation for large-scale learning and complex non-batched scenes on CPUs. Genesis AI's strategy centers on evaluation as the bottleneck for scalable robotics, aiming for physical AI that improves at the speed of compute. Their simulation platform already achieves speeds two orders of magnitude faster than real hardware, with high correlation to on-hardware results.

Requirements

A physicist and an engineer at once.
Ability to judge a method by its production readiness at real scale and ensure it holds up.
Relentless pursuit of even seemingly insignificant defects.
Strong background in physics-based simulation, preferably related to robotics: RBD, FEM, MPM, SPH, IPC, XPBD, VBD, ABD.
Experience with constrained optimization and numerical integration of stiff systems.
A track record of shipping simulation code that others rely on (in an engine, in industry, or in a research codebase).
Solid HPC programming (CPU and/or GPU).
An instinct for what makes a numerical method fast in practice, beyond complexity classes.
Rigor in validation: analytical closed forms, cross-engine consistency, real-world data.

Nice To Haves

Publications in simulation, graphics, or robotics venues (SIGGRAPH, ICRA, IROS, CoRL, RSS).
Contributions to an open-source physics engine.

Responsibilities

Push the physics of Genesis-World forward.
Ship production-ready simulation capabilities that matter for the company's internal needs.
Conduct research applied end-to-end, from algorithm to merged, tested, documented code that real robot-learning pipelines depend on.
Make the engine measurably better along five axes: Speed, Completeness, Fidelity, Versatility, and Scalability.
Develop algorithms that are faster and smarter, spending compute only where it matters (larger stable timesteps, selective fidelity, structure-aware solvers).
Expand the engine to cover new physics domains such as water, human animation, air flow, gravel, tendons, and body organs.
Improve realism of models for contact, friction, deformation, energy, actuation, and materials.
Develop extensible multi-physics without compromising realism, allowing all solvers in the scene to be coupled together at once.
Enable users to write their own solvers that integrate seamlessly into the scene, fostering an open solver ecosystem.
Ensure scalability from workstation to city scale, handling thousands of interacting entities without losing physical soundness.
Establish Genesis-World as the go-to simulator for physical AI.
Invent physics level-of-detail (LOD) to simulate at full fidelity what agents interact with and coarsely what they do not.
Support heterogeneous environments where each parallel world can hold a different model.
Implement adaptive timesteps per island using error-based control and Time-of-Impact stepping.
Develop efficient and accurate two-way constraint-based coupling between heterogeneous grey-box solvers.
Improve scalability of rigid constraint solvers.
Unify contact resolution for hydro-elastic compliance, unilateral constraints, and sequential impulses.
Handle closed kinematic loops intrinsically for numerical stability and speed.
Write physics in plain Python and leverage Quadrants for performance on all backends.
Validate physics through analytical closed forms, other engines, and real-world data.