Generative Codec Avatars Research Scientist

Develop and apply state-of-the-art generative models, such as Diffusion Models, GANs, and VAEs, to build foundational human-centric models for codec avatars.

• Design and run large-scale experiments to evaluate the performance of generative models on various tasks, including:
  1. Image/video generation: generating high-quality images and videos
  2. Multi-view generation: generating 3D consistent and coherent images and 3D content.
• Collaborate with other teams to integrate generative models into the codec avatar pipeline and push the research that enables mid- to long-term products.
• Publish research results in top-tier journals and at leading international conferences.

Minimum Qualifications

• Currently has, or is in the process of obtaining a Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience. Degree must be completed prior to joining Meta.
• Currently has or is in the process of obtaining a PhD degree and/or postdoctoral assignment in the field of computer vision, computer graphics, machine learning, or a related field
• First-author publications in computer vision, machine learning, or computer graphics peer-reviewed conferences (e.g., CVPR, ECCV, ICCV, NeurIPS, ICLR, or SIGGRAPH, etc.)
• Experience with generative models, such as Diffusion Models, GANs, or VAEs, and/or Vision Transformers

Preferred Qualifications

• Proven track record of achieving significant research results as demonstrated by grants, fellowships, and/or patents
• Demonstrated software engineer experience via an internship, work experience, coding competitions, or widely used contributions in open-source repositories (e.g., GitHub)
• 3+ years of experience with prototyping algorithms in Python
• Experience in implementation of large-scale training: parallelisms, sharding strategies, custom kernels.
• Experience with large-scale experiment design and execution.
• Familiarity with the avatars pipeline and/or experience working with human-centric data