AI Engineer - Simulation Environments

Description

You would join the early AI team as a key member, working directly with AI leadership and researchers to define the simulated worlds where the company's humanoid robots learn and get evaluated.

The environments you build sit between research, hardware and product, and their realism and usefulness determine how quickly and safely new robot capabilities reach the field.

The company is an early-stage European robotics business. Its leadership has built large companies before and has experience with robotics standards, and it is substantially funded internally and backed by top-tier investors.

The company times its capital to the point where deployment takes off, and aims to become a European leader in a market it expects to reach a trillion dollars.

Responsibilities

• Design and build scalable training and evaluation environments for humanoid robots using physics engines such as MjLab and Isaac Lab.
• Work with AI researchers to create difficult training tasks, curricula and evaluation scenarios.
• Build and maintain the database of robot, object and scene assets, including import, validation, collision and physics properties, metadata and versioning, so environments can be assembled quickly and reproduced exactly.
• Own large-scale simulation benchmarks and evaluation sweeps that give fast policy comparisons, with sound statistical metrics, regression tracking and clear reports for research teams.
• Work with hardware and control engineers to make simulation match the real robot through robot modeling, actuator and sensor models, system identification and checks against real-world data.
• Work with product teams to turn real customer use cases into simulated environments and benchmarks that measure what matters for deployment.

Requirements

• At least 3 years of hands-on experience building simulation environments for robot learning with MuJoCo or MjLab, MuJoCo Warp or MJX, Isaac Lab or Isaac Sim, or Newton.
• Solid understanding of rigid-body dynamics and contact simulation, including solver parameters, numerical stability and the trade-off between simulation speed and fidelity.
• Experience with robot and scene description formats and asset pipelines, including URDF, MJCF and OpenUSD, mesh processing and convex decomposition (such as CoACD or V-HACD), and asset validation at scale.
• Experience with GPU-parallel simulation, sweep orchestration, reproducible experiments and statistically sound metrics.
• Proficiency in Python and C++, with PyTorch and/or JAX, and GPU programming using CUDA or NVIDIA Warp.
• You enjoy the empirical side of AI research and want to speed up the team's research with strong tools, so everyone can make progress faster together.
• You are a curious builder who works well in a collaborative research setting and can turn the needs of research, hardware and product into environments and benchmarks that people trust.
• Plus: experience with video game engines (Unreal Engine, Unity or Godot), including level design, procedural content generation and real-time rendering.
• Plus: experience with sim-to-real techniques such as system identification, actuator modeling, domain randomization and sensor simulation for cameras, depth, IMU and tactile sensors.
• Plus: experience with photorealistic rendering and scene reconstruction, such as ray tracing, Gaussian splatting or NeRFs.
• Plus: experience with 3D content tools such as Blender or Houdini, and large asset sources such as Objaverse.
• Plus: familiarity with reinforcement learning and imitation learning for humanoid locomotion and manipulation.

Conditions

This is a full-time, permanent position.