Описание

A robot can pick up a mug perfectly in simulation and still drop it on the first real try. Before a humanoid robot handles anything in a factory, it has already practiced thousands of times in simulation. If the friction is slightly off or a contact point jitters, the policy learns a trick that only works inside the sim. On real hardware, the mug slips. That's the challenge: building simulations accurate enough that what a robot learns there still works in the real world. We're working with a humanoid robotics company that builds general-purpose humanoid robots and already deploys them in real industrial environments. Their VLA team teaches robots dexterous manipulation for industrial and service settings, and every one of those skills starts in simulation. This is not a classic simulation role. You'll own the full loop, from designing the environment and the reward to checking that trained policies actually behave on the robot, and fixing the sim when they don't.

Responsibilities

Build gym environments for manipulation tasks
Design reward functions
Set up physics (contacts, friction, mass, joint limits)
Validate the sim against real robot data
Make training run faster on GPU clusters

Requirements

3+ years building simulation environments, with shipped products or published work to show
Deep hands-on experience with MuJoCo, Isaac Sim / Isaac Lab, PyBullet, Drake, or physically accurate work in Unreal or Unity
Strong Python: you can profile bottlenecks and debug physics issues
Experience running large-scale parallel simulation on GPUs

Nice to have

RL experience
Reward design experience
Domain randomization experience
Contact-rich manipulation experience

We offer

Salary: ~£185k–250k gross, depending on the level confirmed during interviews
Relocation to London is required
The company covers relocation costs and helps with the visa and paperwork
Stock options
Private healthcare
30+ paid days off

Hiring process

Before each interview stage, from live coding to the final with leadership, we send you a prep guide on what the team checks and how to get ready