Applied Science Intern – World Model

This is a fully remote position, open to applicants in California.

📋 Description

• Investigate the application of World Models to enhance understanding, simulations, and ultimately the generation of sports or eSports matches (e.g., soccer, DOTA).

• Design, develop, and refine AI video generation models, particularly focusing on World Models; experiment with advanced autoregressive architectures.

• Create and implement cutting-edge algorithms for synthesizing sports matches.

• Be prepared for hands-on tasks to support our stables where we gather data for AI horse video models (just a humorous note, but being versatile and possessing a great sense of humor is essential).

• Collaborate closely with other teams on extensive video-action datasets, designing and executing a comprehensive data-cleaning and preprocessing pipeline.

• Establish robust validation strategies and implement tailored evaluation metrics to compare synthetic and real gameplay.

• Keep up-to-date with the latest literature in relevant fields, such as CVPR, NeurIPS, ICML, ICCV, and assist in aligning it with our strategic roadmap.

⛳️ Requirements

• Currently pursuing a PhD! (preferably located in the San Francisco area)

• Published in top-tier Computer Vision, AI, or Graphics conferences (e.g., CVPR, ICML, ICCV, Siggraph, NeurIPS).

• Proven hands-on experience in building and operating generative computer vision models (e.g., GANs, DiT, VAE).

• Strong understanding of various neural architectures and paradigms (e.g., Transformers, Denoising Diffusion Models, RNNs, Sequence Models, CNNs).

• Solid knowledge of Variational Autoencoders (e.g., ELBO).

• Fundamental understanding of Reinforcement Learning.

• Proficient in Python and PyTorch.

🏝️ Benefits

• Competitive salary and performance-based bonuses.

• Flexible working hours and a supportive work environment.

• Opportunities for professional development and continuous learning.

• Engaging and collaborative team culture focused on innovation.