Research Scientist – Hand Tracking & Manipulation

What you'll do

• Develop and train models for 3D hand pose estimation, mesh recovery, and kinematic tracking, including distributed multi-GPU training.
• Build egocentric hand-object interaction models that track objects, segment tools from hands, and estimate contact information.
• Prototype perception systems for emerging needs such as tactile-visual fusion, action segmentation, and hardware sensor integration.

What they ask for

• Deep expertise in deep learning, 3D computer vision, and articulated tracking or hand pose estimation.
• Experience training large-scale vision models from scratch.
• Strong understanding of parametric hand 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 in CVPR, ICCV, ECCV, or NeurIPS on relevant topics.
• Experience with egocentric manipulation datasets or custom CUDA kernels for 3D operations, differentiable rendering, or collision/contact computation.

Pay

$170 – $210,000 (base salary)

Tools

PyTorch, CUDA, CVPR, ICCV, ECCV, NeurIPS, Ego4D, Ego-Exo4D

AI focus

Research and development of foundation models for hand tracking, articulated pose estimation, hand-object interaction, and related perception for embodied AI and robotics.