Founding Research Scientist - Multimodal Robotics

Adexis; Greater London; Sur place; GBP 150 000 - 225 000; Plein temps

Avantages offerts par ce poste

Salary £150-225k

Equity 0.5–2.5%

Private healthcare (AXA)

Top-of-the-line equipment

Résumé du poste

Adexis, a London-based frontier research lab, is hiring a Member of Technical Staff - Research to lead end-to-end work on visuo-tactile foundation models and dexterous manipulation.

You will own architecture, fusion of vision and touch, and the post-training stack to deploy policies. This is a founding role with equity and a competitive compensation package.

Bring experience with large multimodal models, imitation learning, and real-world robotics; teleoperation and real-world RL are a plus.

Qualifications

• Experience training large multimodal or sequence models from scratch.
• Ability to design data mixes, curricula, and debugging workflows.
• Strong intuition for multimodal fusion and dexterous manipulation.
• Experience transferring policies from simulation to real hardware.

Responsabilités

• Own the model architecture for visual and tactile signals, including tokenization and fusion.
• Make the hard representational calls: define contact signals and time alignment between modalities.
• Own the post-training stack to convert a pretrained model into a deployable policy.
• Design the recipe: data collection strategy, imitation and reinforcement or self-improvement stages.
• Bridge sim-to-real and human-to-robot gaps to ensure policies survive real hardware.
• Define evaluation criteria to detect policies that look good offline but fail on hardware.

Connaissances

Multimodal models

Sequence models

Policy learning / RL

Imitation learning

Description du poste

Adexis, a London-based frontier research lab, is hiring a Member of Technical Staff - Research to lead end-to-end work on visuo-tactile foundation models and dexterous manipulation.

You will own architecture, fusion of vision and touch, and the post-training stack to deploy policies. This is a founding role with equity and a competitive compensation package.

Bring experience with large multimodal models, imitation learning, and real-world robotics; teleoperation and real-world RL are a plus.