Technical Lead Manager, Physical AI

About the role

You will own the end to end lifecycle of Physical AI systems, translating high level research ambitions into scalable production pipelines that power robot learning and autonomous decision making. This role requires you to spend the majority of your time deeply coding and designing experiments while simultaneously providing strategic direction for a small research team. You will be responsible for defining how large foundation models perceive, reason, and act in physical environments, ensuring that the underlying data and training methodologies generalize across diverse embodiments. You will act as the primary interface between cutting edge research and deployable products, ensuring that insights from top conferences are rapidly integrated into the platform. A significant portion of your impact will come from mentoring engineers and researchers, elevating the quality of experimentation, and maintaining rigorous standards for evaluation. You will collaborate closely with data labeling, product, and operations partners to ensure that the data infrastructure supports the full training lifecycle. Ultimately, your work will determine how reliably real world robots can learn new skills and adapt to new settings using the infrastructure built at Scale.

Key facts

What you'll do

• Design and execute a research program around scaling laws for Physical AI, defining data regimes, model sizes, and training schedules that maximize generalization for embodied agents.
• Architect and evaluate Vision Language Action models and world model frameworks, creating new benchmarks that reflect real robot constraints and capabilities.
• Implement production grade code for training, fine tuning, and deploying state of the art models, ensuring that research prototypes can scale to massive datasets.
• Build and maintain robotic native data pipelines in partnership with labeling teams, leveraging VLMs for automated trajectory annotation, synthesis, and quality assurance.
• Conduct experiments in data collection, cross embodiment training, and policy fine tuning to improve sample efficiency and robustness across different hardware platforms.
• Engage directly with customers to understand how they deploy Scale data, translating their requirements into technical specifications for the Physical AI stack.
• Mentor and develop a team of elite researchers, fostering a culture of rapid experimentation, reproducibility, and rigorous empirical analysis.
• Translate findings from top tier conferences such as NeurIPS, ICRA, and CVPR into production ready modules that enhance the capabilities of Scale partners.
• Coordinate with Product and Operations teams to move research prototypes through the full stack into deployed features with measurable impact.
• Define and track key performance indicators for model behavior, safety, and efficiency, ensuring that Physical AI systems meet stringent real world standards.
• Drive the creation of internal tools for experiment management, data versioning, and model monitoring to streamline the research to deployment cycle.
• Establish evaluation protocols that combine simulation based testing with real world trials to validate generalization across tasks and environments.
• Identify failure modes in current models, design targeted experiments to probe weaknesses, and iterate on architectural and training improvements.
• Represent Scale AI in collaborative research initiatives, contributing to open benchmarks and shaping the direction of Physical AI as a field.

Requirements

• Demonstrate expert level proficiency in PyTorch, including a deep understanding of how to optimize training and inference for large scale models.
• Show advanced knowledge of Transformer architectures, attention mechanisms, and self supervised learning techniques used in multimodal settings.
• Provide evidence of hands on experience with diffusion models for sequence generation or for building generative world models that support predictive reasoning.
• Exhibit strong competence in the Physical AI stack, including imitation learning, reinforcement learning, and multi modal sensor fusion strategies.
• Prove experience running large scale distributed training on GPU clusters, with fluency in high performance data loading and pipeline parallelism.
• Have at least one year of experience leading technical teams or projects in a research intensive environment, with clear outcomes to show for it.
• Hold a strong academic or industry background in machine learning, robotics, or related fields, with the ability to read and implement complex research papers.
• Communicate effectively with both technical and non technical stakeholders, translating deep research concepts into actionable plans and clear tradeoff discussions.

Nice to have

• Include first author publications from top tier venues such as NeurIPS, CVPR, ICRA, or CoRL that demonstrate impact in Physical AI.
• Show experience transferring models across different robot form factors, including robotic arms, mobile bases, and humanoid platforms.
• Describe familiarity with high fidelity simulators such as Isaac Gym or MuJoCo, and detail work on sim to real techniques that reduce deployment risk.

Practical notes

• This is a full time position based in San Francisco, Canada.
• Compensation includes base salary, equity subject to Board approval, and a comprehensive benefits package.
• Benefits encompass health, dental, and vision coverage, retirement plans, learning and development stipends, generous paid time off, and potential commuter stipends.
• Candidates who have previously applied for this role must wait 90 days before reapplying.
• Reasonable accommodations are available for applicants with disabilities; interested candidates may contact accommodations@scale.com for support.