Staff Engineer, World Models Data & Evaluation

Orbifold AI; Palo Alto (CA); On-site; USD 180,000 - 260,000; Full time

Job summary

Orbifold AI is hiring Members of Technical Staff to build the data and evaluation foundations for world models and embodied AI systems. The role emphasizes real-world performance and practical impact, not solely theoretical work.

Candidates will design data pipelines, evaluation frameworks, and scalable experiments with robotics partners. You will collaborate with internal teams and external researchers to translate failures into data curation and training cycles, shaping how models learn from

Qualifications

• Self-driven with experience in fast-paced applied research or startups.
• Experience in robotics, embodied AI, world models, and multimodal reasoning.
• Understanding of model training and RL workflows at scale.
• Experience designing data pipelines and evaluation frameworks.

Responsibilities

• Co-design data pipelines and training workflows end to end with robotics and world model teams.
• Define multimodal data structure and indexing for training and evaluation.
• Build scalable data ingestion, cleaning, annotation, and curation systems.
• Work on training or fine-tuning models for perception and multimodal reasoning.
• Iterate datasets and training setups based on model performance, linking eval findings to data curation.

Skills

Applied ML systems

Robotics experience

World models

Multimodal reasoning

Reinforcement learning

Data pipelines

Evaluation frameworks

Production-grade systems

Embodied AI knowledge

Job description

Orbifold AI is hiring Members of Technical Staff to build the data and evaluation foundations for world models and embodied AI systems. The role emphasizes real-world performance and practical impact, not solely theoretical work.

Candidates will design data pipelines, evaluation frameworks, and scalable experiments with robotics partners. You will collaborate with internal teams and external researchers to translate failures into data curation and training cycles, shaping how models learn from