ML Engineer - Simulation & RoboticsUp to £100,000

About the Role

Vsim builds GPU physics simulation and training infrastructure for robot brains, and this role owns the simulation side of their robot learning pipeline. The aim is to cut the team's dependence on physical robot evaluations: realistic, scalable simulations, large-scale synthetic training data, and sim-to-real workflows that make learned behaviours hold up on real hardware. You'll be responsible for the performance and reliability of those learned behaviours on the physical robots themselves.

Day to day you'll train policies with reinforcement learning, imitation learning and related approaches, build out domain randomisation, system identification and robustness-driven training, deploy policies to physical robots and diagnose where simulated and real behaviour diverge. You'll also build the evaluation methods and tooling that measure policy performance and robustness, working closely with the wider engineering team to integrate what you train into their robotic systems. It's a small, experienced team with direct access to advanced robot hardware and edge compute, so you'll have significant technical ownership from the start.

What They're Looking For

• You've built and evaluated learning-based robotics systems, ideally training policies for robotic or embodied systems
• You have strong practical foundations in reinforcement learning, imitation learning or policy learning, with PyTorch or an equivalent framework
• You're a strong Python engineer who has developed, debugged and extended ML or robotics systems, not just research notebooks
• You've worked in at least one of robotics simulation, sim-to-real, domain randomisation, system identification or robustness-driven training
• You can debug across the whole stack, from learning and simulation through software to physical robot behaviour, and pick up unfamiliar technical areas quickly
• Deploying and iterating on learned policies on physical robots is a strong plus, as is an MS or PhD in CS, Robotics or a closely related field

About Vsim

Vsim builds GPU physics simulation and training infrastructure for robotics. It was founded in 2022 by two engineers who each spent over a decade building physics engines on Nvidia's simulation team. VC-backed with a long runway, it's a team of around 11 in Cambridge, UK, most of them engineers.