Robot Learning Scientist: World Models & Policies

Spaitial Ltd.; Greater London; On-site; GBP 90,000 - 130,000; Full time

Job summary

SpAItial Ltd. seeks a senior Research Scientist to train policies that turn world models into capable robots. You will own VLA and WAM pipelines from data to running policy, shaping the embodied research direction and closing the sim-to-real gap through domain randomization and calibration.

You will adapt VLM backbones for control, weight diverse robot data, and design action representations. Strong publications and end-to-end policy training experience are essential.

Qualifications

• PhD in robotics, ML, or computer vision with robot learning focus.
• Publications in top venues or open-source work is a plus.
• Experience training manipulation policies end to end.

Responsibilities

• Own the end-to-end training pipeline for VLA and WAM policies.
• Define technical direction for embodied research at SpAItial.
• Bridge sim-to-real gaps with domain randomization and calibration.
• Adapt VLM backbones for control and tokenization strategies.
• Curate diverse robot datasets and design action representations.
• Lead post-training refinement via supervised fine-tuning and RL.

Skills

Robot policy design

Imitation learning

VLM backbones

Python

PyTorch

Education

PhD in robotics/ML/CV with robot learning focus

Tools

PyTorch

Python

Distributed training (FSDP)

Job description

SpAItial Ltd. seeks a senior Research Scientist to train policies that turn world models into capable robots. You will own VLA and WAM pipelines from data to running policy, shaping the embodied research direction and closing the sim-to-real gap through domain randomization and calibration.

You will adapt VLM backbones for control, weight diverse robot data, and design action representations. Strong publications and end-to-end policy training experience are essential.