Job Description

We are seeking a hands-on [Research Engineer/Research Scientist] with expertise in world modeling, video generation, predictive modeling, or learned simulation. You will develop models that generate environments and future observations, capture temporal dynamics, and respond to actions or interventions.

You will own research problems from hypothesis and implementation through training, evaluation, and failure analysis. Working with colleagues across modeling, data, evaluation, and infrastructure, you will translate research ideas into working models and measurable improvements.

About the Team

Our team has a background in autonomous driving, where understanding dynamic scenes, anticipating interactions, and evaluating the consequences of actions are central challenges. This experience shapes our research into how models can learn to represent, generate, and predict the physical world.

Building on this foundation, we develop learned world models for world generation, future prediction, and simulation. Our work combines video generation with representations of scene structure, motion, and interactions to model how environments evolve under different actions and conditions. We aim to create controllable, coherent simulations that support autonomous driving research and development.

Our research spans model architecture, training methods, data, and evaluation, with a focus on long-horizon consistency, multi-view coherence, and physically plausible dynamics. We bring together modeling and engineering to turn new ideas into working systems and measurable advances in both visual quality and predictive behavior.

Key Responsibilities

• Research, develop, and adapt models for world generation, future prediction, and simulation.
• Advance model capabilities in controllability, consistency, and realistic world dynamics.
• Develop scalable and efficient training and inference systems for world models.
• Improve data and evaluation methods to support model quality and generalization.
• Design experiments, analyze model behavior, and use findings to guide further research.
• Collaborate across research and engineering to turn new ideas into reliable implementations and share technical findings.

Requirements

• Master's degree or PhD in Computer Science, Machine Learning, Electrical Engineering, Robotics, Applied Mathematics, or a related field, or equivalent practical experience.
• Strong programming skills in Python and practical experience with PyTorch.
• Hands-on experience developing and training models in one or more of world modeling, video generation, spatiotemporal prediction, or learned simulation.
• A solid foundation in deep learning and generative or predictive modeling, with the ability to implement and evaluate research ideas.
• Experience with large-scale model training or inference and distributed machine-learning systems.
• Ability to design rigorous experiments, analyze results, and draw sound technical conclusions.
• Strong software-engineering practices, clear technical communication, and collaborative problem-solving skills.

Preferred Qualifications

Experience in one or more of the following areas would be an advantage:

• Research in controllable generation, long-horizon prediction, or interactive simulation.
• Fine-tuning, adaptation, or extension of advanced pretrained video generation and world models, including Cosmos, Wan, LingBot, or comparable models.
• Methods for improving model quality, efficiency, or scalability, including adaptation and distillation.
• Research contributions demonstrated through publications, open-source work, or industry projects.
• Application experience in robotics, embodied AI, autonomous driving, or other simulation domains.