Research Scientist: 3D Hand Tracking & HOI Architect

Kindredventures; New York (NY); On-site; USD 120,000 - 180,000; Full time

Job summary

Mecka AI is seeking a Research Scientist to architect and train proprietary foundations for 3D hand tracking and articulated pose estimation. You will lead development of end-to-end hand-object perception, using proprietary data to surpass public baselines and drive robotics workloads.

The role emphasizes large-scale distributed training, novel loss functions, and integration of perception with downstream robotics policies, in a fast-paced AI hardware environment.

Qualifications

• Deep learning, 3D vision, and articulated tracking expertise.
• Experience building large-scale vision models from scratch.
• Strong knowledge of hand models, IK, and dense mesh estimation.
• Proficiency with PyTorch and scaling frameworks.
• Experience handling massive image/video datasets.

Responsibilities

• Architect and train proprietary 3D hand tracking models.
• Scale training across multi-GPU clusters for large datasets.
• Develop HOI and dense mesh recovery architectures.
• Probe biomechanical constraints and temporal stability.
• Prototype solutions for egocentric manipulation tasks.

Skills

Deep Learning

3D Computer Vision

Articulated Tracking

Hand Pose Estimation

PyTorch

Large-Scale Training

Tools

CUDA

Differentiable Rendering

Multi-GPU Frameworks

Job description

Mecka AI is seeking a Research Scientist to architect and train proprietary foundations for 3D hand tracking and articulated pose estimation. You will lead development of end-to-end hand-object perception, using proprietary data to surpass public baselines and drive robotics workloads.

The role emphasizes large-scale distributed training, novel loss functions, and integration of perception with downstream robotics policies, in a fast-paced AI hardware environment.