Senior RL Engineer - Real-World Robotics (London On-site)

hirehire; Greater London; On-site; GBP 120,000 - 180,000; Full time

Benefits offered by this job

Equity: stock options with meaningful+

23+ days annual leave incl. UK bankhol

Private healthcare

Pension: 8% total contribution

Free meals in the office

Visa sponsorship and relocation to-Ldn

Job summary

Humanoid in London is seeking a senior Reinforcement Learning Engineer to develop language- and vision-conditioned manipulation policies using RL, both in simulation and on real robots. This is an individual contributor role, with no management track, focused on turning a base vision-language-action model into robust robotic policies.

The team operates in a research-adjacent environment where results are shared as blog posts rather than papers.

Qualifications

• 3+ years building deep learning systems in industry or research.
• Proven RL experience solving real problems with neural networks.
• Experience training models (architecture, training and inference) on LLMs, VLMs, or image/video models.
• Strong Python and PyTorch or JAX; able to profile, debug numerics and write maintainable research code.
• Solid coding fundamentals; comfortable with LeetCode-style rounds.
• Proactive, able to tackle open problems and document experiments clearly.
• Working English.

Responsibilities

• Train language- and vision-conditioned manipulation policies with RL, in simulation and in the real world.
• Build challenging and diverse suites of manipulation tasks in simulation.
• Work with Teleoperations to collect trajectories in simulation for behaviour cloning.
• Work with Testing and Operations to set up real-world RL training pipelines.
• Find ways to bring policies trained in simulation to the real world (sim-to-real).

Skills

Deep learning systems

Python programming

PyTorch

JAX

English proficiency

Tools

PyTorch

JAX

Python

Job description

Humanoid in London is seeking a senior Reinforcement Learning Engineer to develop language- and vision-conditioned manipulation policies using RL, both in simulation and on real robots. This is an individual contributor role, with no management track, focused on turning a base vision-language-action model into robust robotic policies.

The team operates in a research-adjacent environment where results are shared as blog posts rather than papers.