[UG/MS] Recruiting Students for World Model / Video Diffusion Research

The following announcement is from Professor Yue Wang. Please contact them directly if you have any questions.

Dear all,

We are conducting a research project on building a world model using video diffusion, leveraging the Cosmos backbone. This project focuses on training a video diffusion model capable of generating high-quality, physically grounded video predictions, with potential applications in autonomous driving and vision-language-action models. This is a collaborative effort with Bosch Research, and we do not rule out submitting our work to top-tier venues in the future.

We are currently recruiting student researchers to assist with:

• Data processing & preparation: Cleaning, organizing, and preprocessing large-scale video datasets for training. (Much of the data pipeline is already in place — you will be guided through each step.)
• 3D vision pipeline: Running point cloud projection and related geometric processing to prepare structured inputs for the model.
• Model training: Training and iterating on a video diffusion model built on the Cosmos backbone, using cloud compute (Vast.ai) and collaborative tools (Dropbox).

What you will gain:

• Hands-on experience in cutting-edge generative model research (video diffusion, world models).
• The opportunity to work alongside researchers from Bosch Research and contribute to a real product development pipeline.
• Practical skills in large-scale model training, cloud computing, and data engineering that will benefit your future research and career.
• Potential co-authorship on future publications, depending on contribution.

We are looking for students who:

• Are interested in generative models, computer vision, or deep learning (experience in these fields is a plus).
• Have basic programming skills in Python and familiarity with PyTorch or similar deep learning frameworks.
• Are self-motivated, detail-oriented, and willing to dedicate consistent time to the project.
• Experience with diffusion models, video generation, or 3D vision.

The project is remote-friendly and you will receive direct, step-by-step mentorship throughout the process.

If you are interested, please fill out the form: https://forms.gle/21Zco6sTigBwDeNw7. There is no need to send a separate email and we will reach out if we find a good fit.

Best,
Yue Wang

Published on March 12th, 2026Last updated on March 12th, 2026