Frontier AI Reality Evaluation Scientist

Movendi Inc.; San Francisco (CA); Hybrid; USD 100,000 - 180,000; Full time

Job summary

Movendi Inc. seeks an evaluation researcher to design novel benchmarks and assess model understanding of reality. You will create experiments around object permanence, long-horizon reasoning, and physical plausibility, pushing frontier lab research forward.

The role emphasizes rigorous evaluation, transparent reporting, and collaboration with researchers to refine benchmarks and interpret results for real-world impact.

Qualifications

• Must demonstrate ability to design and run cognitive benchmarks.
• Proven capability to reason about predictive validity and experimental controls.
• Experience with baseline and evaluation datasets for AI/ ML models.

Responsibilities

• Create entirely new benchmarks and evaluation protocols.
• Assess model understanding of physical and temporal concepts.
• Document evaluation methodology and results for peer review.

Skills

Object permanence

Counterfactual reasoning

Occlusion recovery

Long-horizon consistency

Physical plausibility

Job description

Movendi Inc. seeks an evaluation researcher to design novel benchmarks and assess model understanding of reality. You will create experiments around object permanence, long-horizon reasoning, and physical plausibility, pushing frontier lab research forward.

The role emphasizes rigorous evaluation, transparent reporting, and collaboration with researchers to refine benchmarks and interpret results for real-world impact.