Robotics Applied Research Lead

present of AI; Greater London; Hybrid; GBP 90,000 - 140,000; Full time

Job summary

Humanoid is seeking a researcher-lead to set the research direction for robot learning while doing hands-on model work at least half the year. The role requires driving architecture and training decisions on substantial Transformer systems and guiding a small team to translate customer needs into testable requirements.

You will work at the intersection of offline evaluation and real-world robot testing, gaining exposure to vision-language-action models and multimodal Transformers while building

Qualifications

• Experience leading research in robotics/robot learning.
• Strong background in Transformer-based models.
• Ability to translate customer needs into testable requirements.

Responsibilities

• Set the research direction for robot learning and lead hands-on model work at least 50% of the time in the first year.
• Architect and train Transformer systems; guide a small team.
• Design honest evaluations and validate with both offline and real-world tests.

Job description

This role asks for someone who can set the research direction for robot learning and still do the hands on model work, at least half the time in the first year. A strong candidate has personally made architecture and training calls on substantial Transformer systems and has taken at least one project from offline data into a real physical system. They also need to build a small team, design honest evaluations, and turn customer needs into testable requirements.

How it fits the AI era

The work sits at the centre of the AI era: choosing which models to adapt, which data actually improves robot behaviour, and which offline measures predict real world results. AI speeds up training and experimentation, but judgment about what to test, how to spot leakage, and when evidence is strong enough stays with people. Safety limits and staged physical trials also remain human decisions.

How it builds AI experience

Someone here would gain direct experience with vision language action models, multimodal Transformers, and robot foundation models, plus the data and evaluation craft that makes them work on hardware. Building that background means hands on training and debugging of Transformer models, careful study of how data quality and mix change outcomes, and real robot testing rather than simulation alone.

Worth showing

• Offline and real world evaluation
• Python and PyTorch fluency

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Humanoid

Humanoid builds humanoid robots for industrial work, paired with its KinetIQ AI framework (per the company).