Want to know if this job is worth applying to?
San Francisco, California, United States
Onsite
Physical Intelligence is bringing general-purpose AI into the physical world. We are a group of engineers, scientists, roboticists, and company builders developing foundation models and learning algorithms to power the robots of today and the physically-actuated devices of the future.
High-quality data is the fuel for our work. The Annotations team produces high-quality data labels to support model development.
This role is a hands-on owner for a portfolio of annotation programs: partnering with researchers to understand what data they need and why, translating that into concrete requirements and rubrics, and running the operation that delivers the data. This person will also be expected to drive continuous improvement across multiple surfaces to mature the annotations function and balance strategic problem solving with tactical execution.
Own end-to-end performance for a portfolio of annotation programs, driving quality, speed, and efficiency
Partner with researchers to define and refine requirements
Drive step-function continuous improvement in our operations, identifying issues and building scalable solutions
Foster clarity, collaboration, and alignment across both technical and nontechnical stakeholders, including vendor partners, researchers, operations
5+ years of experience in program management, operations, technical project management, consulting, startup operations, or similar roles
Strong problem solving and systems thinking; ability to identify points of leverage in a constantly-evolving system and diagnose problems with incomplete information
Intuition for data quality and research needs, grounded in understanding of model training and evaluation
Stellar communication and ability to clarify nuance, structure complex information in digestible ways, and work with remote, ESL audiences
Organizational skills to bring structure to chaos and balance needs across multiple programs
Experience with data annotations for ML/AI, having seen versions of this problem before
Experience with robotics, egocentric video, or other multimodal sensor data
Experience with using coding agents to fill gaps in tooling / infra
