World Model Research Engineer (MOTHER EXO)

Media Stream AI Limited; Manchester; Hybrid; GBP 70,000 - 110,000; Full time

Job summary

Media Stream AI Limited is seeking a research engineer to advance the EXO world model, focusing on latent dynamics and video sequence modeling. You will work with the MOTHER backbone, building evaluation harnesses and translating published work into practical stack components.

The role requires strong ML research engineering and a track record of rigorous publication or reproduction, with hands-on PyTorch on multi-GPU setups. Manchester-based factory, start dates arranged individually.

Qualifications

• Strong ML research engineering background with focus on sequence and video models.
• Experience with latent dynamics and world models.
• Proficiency in PyTorch at multi-GPU scale; proven publication or reproduction track record.

Responsibilities

• Train and evaluate world-model and latent-dynamics heads on the MOTHER backbone.
• Design rollout and evaluation harnesses to measure prediction quality.
• Translate published research into production stack and assess transferability.
• Collaborate with controls to convert predictions into actionable behavior.

Skills

Machine learning research

Video/sequence modeling

Latent dynamics

Research engineering

Tools

PyTorch

Multi-GPU training

Job description

Advance the EXO world model: latent video and action dynamics that let the robot predict the consequence of an action before it takes it.

Key responsibilities:

• Train and evaluate world-model and latent-dynamics heads on the frozen MOTHER backbone.
• Design rollout and evaluation harnesses that measure prediction quality honestly.
• Bring published research into the stack and report what does and does not transfer.
• Work with controls and manipulation to turn prediction into behaviour.

Requirements:

Strong machine learning research engineering; sequence and video models, latent dynamics, VLA or world models; PyTorch at multi-GPU scale; publishes or reproduces rigorously.

Salary:

Salary on application

The Manchester factory comes online Q1 2027. Roles are recruited ahead of that date, with start dates agreed individually.