Embodied AI Policy Evaluation Engineer

Orbifold AI, Inc.; Palo Alto, Northern (CA, KY); Hybrid; USD 180,000 - 230,000; Full time

Job summary

Orbifold AI, Inc. is building the infrastructure layer for Physical AI. This role focuses on evaluating robot policies by running trained policies against thousands of real scenes and generating repeatable failure taxonomies.

You will collaborate with partner research teams, translate evaluation findings into concrete data specifications, and help design end-to-end evaluation methodologies that scale to corpus-level testing.

Qualifications

• PhD or equivalent research background with first-author work.
• Experience training and debugging real policies on robots.
• Strong statistical literacy and evaluation design.
• Proficiency with PyTorch and Ray at scale.

Responsibilities

• Design end-to-end evaluation methodology.
• Build fine-grained failure taxonomies.
• Train and fine-tune reference policies.
• Build automated critics and judges.
• Run evaluations on real hardware.
• Collaborate with partner research teams.
• Translate findings into concrete data specs.

Skills

PhD or equivalent

Robot learning

PyTorch

Ray

First-author pubs

Self-driven

Education

PhD or equivalent

Tools

Ray

OpenVLA

Job description

Orbifold AI, Inc. is building the infrastructure layer for Physical AI. This role focuses on evaluating robot policies by running trained policies against thousands of real scenes and generating repeatable failure taxonomies.

You will collaborate with partner research teams, translate evaluation findings into concrete data specifications, and help design end-to-end evaluation methodologies that scale to corpus-level testing.