Machine Learning Engineer, World Models & Real-Time Simulation

Institute of Foundation Models; Abu Dhabi; On-site; AED 331,000 - 441,000; Full time

Job summary

The Institute of Foundation Models (IFM) at MBZUAI is seeking a Machine Learning Engineer to own the PAN world model backbone. You will design and sustain large‑scale data pipelines, distributed training for diffusion transformers, and game‑engine simulation environments for world‑model integration into robotics.

You will apply PyTorch expertise to debug training/inference, build scalable systems, and collaborate with researchers to translate ideas into reliable production features.

Qualifications

• Degree or higher in ML, CS, Robotics or related field.
• Hands-on with state-of-the-art video generative models and world models (Cosmos-3, LTX 2.3, Self-Forcing, Lingbot-World or equivalents).
• Deep expertise in at least one area: data pipelines, model training, rendering/simulation, or robotics integration.
• Strong engineering in PyTorch and ability to debug training/inference issues.

Responsibilities

• Build and maintain large-scale video and simulation data pipelines for world model training.
• Develop distributed training systems for diffusion transformers and world models.
• Create interactive simulation environments (Unreal Engine, Blueprint gyms) for training and evaluating world models.
• Integrate world action models into robotics systems; optimize end-to-end performance.

Skills

Video models

World models

Diffusion transformers

PyTorch

Robotics integration

Education

BSc or above in ML/CS/Robotics

Tools

Unreal Engine

Blueprint-based engines

Job description

The Institute of Foundation Models (IFM) at MBZUAI is seeking a Machine Learning Engineer to own the PAN world model backbone. You will design and sustain large‑scale data pipelines, distributed training for diffusion transformers, and game‑engine simulation environments for world‑model integration into robotics.

You will apply PyTorch expertise to debug training/inference, build scalable systems, and collaborate with researchers to translate ideas into reliable production features.