Simulation Architect for Physical AI & World Models

Veeda; California (MO); Hybrid; USD 150,000 - 190,000; Full time

Job summary

Veeda AI is seeking a Member of Technical Staff in Simulation to join a fast-moving team focused on multimodal world models for Physical AI. You will connect our world models to popular simulators, rethink harness approaches, and own synthetic data generation to improve training loops.

You will design procedural scenes on OpenUSD/URDF, ensure deterministic rollouts, and drive fidelity between simulators and real hardware while matching cutting-edge research to practical product goals.

Qualifications

• Bachelor's degree or equivalent hands-on experience in CS/CG/Robotics.
• Experience building or extending a physics engine or simulation framework (Newton, MuJoCo, PhysX, or proprietary).
• Strong C++ and Python skills and GPU coding experience (CUDA, NVIDIA Warp).
• Fluency in 3D content pipelines (OpenUSD, glTF, meshes, materials, frames, units, inertias).
• Ability to measure simulation fidelity and close gaps between simulators and hardware.

Responsibilities

• Simulator integration: connect world models to existing simulator APIs (Isaac Lab, Carla).
• Harness engineering: create a physical AI harness analogous to an LLM harness.
• Synthetic data generation: own data collection for success/failure cases.
• Scene & Asset Pipelines: build procedural scene generation and asset authoring on OpenUSD, URDF, MJCF with automated decomposition and inertia assignment.
• Throughput & Determinism: keep rollouts replayable with fixed seeds for regression analysis.
• Environment API & Fidelity: manage interface to policy training runs and regression benchmarks.

Skills

C++

Python

GPU programming

OpenUSD

3D content pipelines

Education

Bachelor's degree or equivalent

Tools

CUDA

NVIDIA Warp

OpenUSD tooling

Job description

Veeda AI is seeking a Member of Technical Staff in Simulation to join a fast-moving team focused on multimodal world models for Physical AI. You will connect our world models to popular simulators, rethink harness approaches, and own synthetic data generation to improve training loops.

You will design procedural scenes on OpenUSD/URDF, ensure deterministic rollouts, and drive fidelity between simulators and real hardware while matching cutting-edge research to practical product goals.