Current opportunities.
The opportunity
Join a forward-thinking autonomy team developing advanced humanoid robots. In this role, you will develop and train reinforcement learning models for robotic manipulation tasks, working both in simulation and real-world environments. Collaborate across teams to create robust manipulation policies, contribute to deployment pipelines, and innovate solutions to bridge simulation and reality performance gaps.
What you could work on
- Train language-vision conditioned manipulation policies using reinforcement learning in simulation and real-world settings
- Design and implement diverse and challenging manipulation tasks in simulation
- Collaborate with teleoperation teams to collect trajectories for behavior cloning
- Work with testing and operations to establish real-world RL training pipelines
- Experiment with methods to transfer policies from simulation to physical robots
Relevant experience
- 3+ years building deep learning systems with delivered models or publications
- Hands-on experience with at least one of: large language models, vision-language models, or image/video generative models
- Experience applying reinforcement learning with deep neural networks in practical domains
- Strong proficiency in Python and PyTorch or JAX, including debugging and profiling numerical code
- Self-driven with clear communication, documentation, and concise presentation of trade-offs
- Desirable: experience with robotics simulators (e.g., Isaac Sim, MuJoCo)
- Desirable: experience in RL for robotics and large-scale RL infrastructure (e.g., ray)
- Desirable: publications at major ML conferences or contributions to open source
Security vetting
Confirm any vetting requirements with your consultant. An absent badge does not mean vetting is unnecessary.
Working arrangements
London - Fully onsite, Full time. Confirm employment terms and any clearance requirements with your consultant.