Research Scientist – Body Pose Detection

What you'll do

• Design and train models for 3D pose estimation, full-body mesh recovery, and kinematic tracking, including large-scale distributed training.
• Develop robust human-scene interaction and motion models for occlusion, dynamic scenes, physical contact, and downstream robotics.
• Prototype perception models for emerging tasks such as action segmentation, intent prediction, and new sensor integrations.

What they ask for

• Deep expertise in deep learning, 3D computer vision, articulated tracking, or human body pose estimation.
• Proven experience training large-scale vision models from scratch.
• Strong understanding of parametric human body models, inverse kinematics, and dense mesh estimation.
• Mastery of PyTorch and deep learning scaling frameworks; experience curating massive, multi-terabyte image and video datasets.

Nice to have

• First-author publications at CVPR, ICCV, ECCV, or NeurIPS on relevant human pose, motion, or interaction topics.
• Experience with human motion datasets such as AMASS, Human3.6M, EgoBody, PROX.

Tools

PyTorch, SMPL, SMPL-X, GHUM, MHR, SOMA-X, AMASS, Human3.6M

AI focus

Builds foundation models for human pose and motion understanding, human-scene interaction, and physics-aware tracking for embodied AI and robotics.