Lead Research Engineer, 3D Perception & Reconstruction

Bonfirevc; Palo Alto (CA); On-site; USD 120,000 - 180,000; Full time

Job summary

Orbifold AI seeks a Research Engineer to advance 3D perception and reconstruction in Palo Alto, on-site. You will own the end-to-end 4D reconstruction stack, including 6-DoF object pose, hand and full-body pose, and multi-view tracking, ensuring accuracy across challenging scenarios.

You will collaborate with partner researchers, push improvements in occlusion and fast motion handling, and train on in-house golden datasets to demonstrate model progress and reliability.

Qualifications

• PhD or equivalent in 3D computer vision or closely related field.
• Depth in pose estimation, multi-view or long-horizon tracking, neural reconstruction (NeRF, 3DGS) or SLAM.
• Experience with real capture rigs and handling failure modes like occlusion, motion blur, desynchronization.

Responsibilities

• Own the 3D reconstruction stack end to end: 6-DoF pose, hand and full-body pose, multi-view tracking.
• Push accuracy under occlusion, self-contact, fast motion and reflective materials.
• Train on in-house golden datasets with motion capture and force sensing; prove model improvements.
• Build automated critics and metrics to catch calibration drift and label noise.
• Define what verified means for each modality and embodiment and maintain standards.
• Collaborate with partner researchers who audit labels and defend numbers.

Skills

PyTorch

Ray

Hands-on research

Self-driven

Education

PhD or equivalent in 3D computer vision

Tools

NeRF

3DGS

SLAM

Camera calibration

Job description

Orbifold AI seeks a Research Engineer to advance 3D perception and reconstruction in Palo Alto, on-site. You will own the end-to-end 4D reconstruction stack, including 6-DoF object pose, hand and full-body pose, and multi-view tracking, ensuring accuracy across challenging scenarios.

You will collaborate with partner researchers, push improvements in occlusion and fast motion handling, and train on in-house golden datasets to demonstrate model progress and reliability.