Robot World Models Engineer: Visual Prediction & Learning

Ashby, Inc.; Cambridge; On-site; GBP 65,000 - 100,000; Full time

Job summary

Dynamic Creatures in Cambridge is seeking a hands-on researcher or engineer to develop visual world models for robots. You will learn from images, video, and robot interaction data to build representations of the physical world and predictive models that guide robot policies.

The role emphasizes self-supervised learning, action-conditioned prediction, and close collaboration with robotics and controls engineers to apply research to real-world robots.

Qualifications

• Hands-on experience with world models or video representation learning.
• Strong Python and PyTorch skills with ownership of model training and evaluation.
• Solid foundation in computer vision and temporal/sequential modelling.
• Ability to run rigorous experiments with clear baselines and evaluations.

Responsibilities

• Build and evaluate visual world models that capture scene dynamics and object interactions.
• Develop models from video and robot data, including self-supervised learning and action-conditioned prediction.
• Connect learned representations to planning and robot behaviour in collaboration with engineers.
• Own experiments from data collection to training, evaluation and validation on physical systems.
• Design evaluations testing predictive accuracy, generalisation, robustness and usefulness for downstream tasks.
• Investigate failures and address inference latency and compute constraints.

Skills

Python

PyTorch

Computer vision

Experimentation

Temporal modelling

Tools

PyTorch

OpenCV

NumPy

Job description

Dynamic Creatures in Cambridge is seeking a hands-on researcher or engineer to develop visual world models for robots. You will learn from images, video, and robot interaction data to build representations of the physical world and predictive models that guide robot policies.

The role emphasizes self-supervised learning, action-conditioned prediction, and close collaboration with robotics and controls engineers to apply research to real-world robots.