At a Glance
- Tasks: Transform foundation models into dexterous robotic behaviour and design learning strategies.
- Company: Join Adexis, a cutting-edge research lab in London focused on Physical AI.
- Benefits: Competitive salary, equity options, and a vibrant team culture.
- Other info: Dynamic environment with opportunities to work alongside top-tier researchers.
- Why this job: Be part of a pioneering team solving the toughest challenges in embodied intelligence.
- Qualifications: Experience in training manipulation policies and expertise in imitation and reinforcement learning.
The predicted salary is between 150000 - 225000 £ per year.
Adexis is a London-based frontier research lab building the somatosensory cortex for Physical AI to solve dexterous manipulation. Backed by leading VCs, the team includes researchers from Anthropic, ElevenLabs, and Google DeepMind, led by a CEO with a track record of building production ML at Tractable and Wayve.
As a founding hire at Adexis, you will own the post-training recipe that transforms foundation models into reliable, dexterous robotic behaviour. You will bridge the gap between world representations and physical execution, designing the imitation and reinforcement learning strategies that allow robots to feel and manipulate the world with human-level precision.
Location: London, UK
Why this role is remarkable:
- Join a high-density founding team of outlier researchers from Anthropic, Google DeepMind, and ElevenLabs to solve the hardest problem in embodied intelligence.
- Own the critical technical boundary between world models and physical action, defining how machines learn the power of touch and dexterous manipulation.
- Benefit from a talent-dense culture inspired by Bell Labs and Skunkworks, supported by top-tier VCs and a leadership team with deep experience scaling AI.
What You Will Do:
- Architect the entire post-training stack, converting pretrained foundation models into dexterous control policies using teleoperated demonstration data and imitation learning.
- Close the sim-to-real and human-to-robot gaps, ensuring policies learned from demonstrations survive contact and maintain reliability on physical hardware.
- Design and implement comprehensive evaluation frameworks to catch policy failures that are invisible offline but catastrophic in real-world contact-rich environments.
The ideal candidate:
- Proven experience training manipulation policies and deploying them on real hardware, with a deep understanding of why policies fail upon contact.
- Deep expertise in imitation learning and reinforcement learning, with the judgment to determine when RL is necessary despite its inherent training instability.
- A strong technical perspective on learning dexterous, contact-rich skills and experience with teleoperation systems or learning from human-led demonstrations.
Member of Technical Staff – Post-Training (£150k-£225k + Equity) at Adexis in London employer: Jack & Jill
At Jack & Jill, we pride ourselves on being an exceptional employer that fosters a dynamic and innovative work culture. Our team enjoys a range of benefits including flexible working arrangements, professional development opportunities, and a collaborative environment that encourages creativity and growth. Located in a vibrant area, we offer unique advantages such as access to cutting-edge technology and the chance to work with industry leaders in AI-driven marketing strategies.
