Codec Avatars Research Engineer

Description:

Meta Reality Lab’s Codec Avatar Research team is building technology to enable immersive, photorealistic social presence. Codec Avatars are real-time live-drivable representations that match the appearance of their users. As part of the Lab’s Instant Codec Avatar group, you’ll work to scale up Codec Avatar technology by modeling the manifold of human motion and applying that model to the process of motion and behavior synthesis.This role is primarily centered on the animation of our codec avatars from multimodal inputs such as audio, image, and text. This involves training large-scale universal prior models using large-scale 2D and 3D datasets, and then applying these models to downstream tasks focusing on agentic Codec Avatars. As part of this role, you will collaborate with a team of research scientists, research engineers, and software engineers to pioneer research in the field of digital humans. You will work with some of the largest 3D motion datasets in the world, supported by an advanced cluster of thousands of GPUs. You will be responsible for designing and conducting research experiments, contributing significantly to our mission of building fully autonomous avatars.

Codec Avatars Research Engineer Responsibilities:
 

• Develop a large-scale machine learning model to build a universal prior for human motion and behavior
• Define data requirements and collaborate with the Research SuperCluster team to efficiently load data
• Partner with FAIR to explore and implement various state-of-the-art models and methods, including transformers, diffusion models, and LLMs
• Collaborate with the Encoder team to support the real-time animation of the avatars
• Define, design, and implement automatic pipeline to leverage large amounts of 2D video data to enhance 3D motion generation
   

Minimum Qualifications:
 

• Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience
• Masters or higher degree in Computer Science or a related technical field, or equivalent experience
• Experience with multi-node machine learning training workflows and frameworks
• Experience in developing or applying computer graphics algorithms
• Experience in developing or applying computer vision algorithms
• Adept communication skills, including experience in driving decision-making processes
• Experience working with cross-functional teams including, research, software, legal, privacy, and security
• Proficiency in Python
• Experience in Linux/shell scripting development
   

Preferred Qualifications:
 

• PhD degree in Computer Science or a related technical field, or equivalent experience
• 5+ years experience in machine learning and modeling of 3D humans
• Publish papers on CVPR, ECCV, ICCV, T-PAMI, SIGGRAPH, SIGGRAPH Asia, etc. as the first author
• Understanding of neural network training, experience with large-scale models, diffusion models, hypernetworks, etc
• Proficiency in C++