Physical AI Resident: Robot Learning & Deployment

Everyone selling a robot has a polished demo and inflated metrics. Almost none have proven performance data from real industrial operations.

Today, buyers are forced to run months-long pilots with every vendor or choose whichever robot looks the best on a presentation.

Deplace is the deployment layer for Physical AI, based on evidence-backed results. We deploy robots on real customer workflows, capture every trial, and continuously evaluate performance using AI-assisted, human-verified measurements, resulting in faster Physical AI development cycles and successful deployments.

We are a VC-funded, well-capitalized startup working at the intersection of Physical AI research and real-world robotics deployment.

In this role, you’ll be working with robot policies such as π0.5 and GR00T on multiple hardware platforms and stressing them on real customer task requirements. This requires being comfortable with the full robotics stack.

You will actively be involved in shaping the direction of the company and its products.

• Own the full pipeline from data collection on real robots to policy training, evaluation, and deployment.
• Actively detect breaking points, debug failures, and model them into insights and data.
• Build data collection tooling for teleoperated demonstrations, apply imitation learning, and process multimodal data.
• Write production-quality code that is traceable, reproducible and modular.
• Brainstorm with us as we grow the company. Helping drive strategy is worth more than closing tickets.

• You have worked on training, evaluating, and deploying a manipulation policy on hardware.
• Strong foundation in Python, ROS, machine learning, and robotics fundamentals: manipulation, perception, and controls.
• A deployed system, publication, thesis or a personal/open-source project. A flagship you can speak about for hours.
• High agency. Ability to break down problems and make autonomous decisions, and know when to seek guidance.

Ideal candidates have three or more of the following.

• Hands-on deployment and operation of bimanual arms, humanoids, or mobile manipulators
• Experience with teleoperation rigs and demonstration collection at volume
• Familiarity with USD scenes and Gaussian splatting for creating sim-ready environments in Isaac Sim, MuJoCo, ManiSkill, or similar
• Experience with multimodal datasets, LeRobot, or comparable tooling
• Owned a large subsystem through FIRST Robotics, BattleBots, Formula Student, or a comparable project.
• Statistical or mathematical background and prior work on theoretical foundations
• Publications or presentations at top outlets like ICRA, CVPR, IROS, CoRL, or similar

• Location: San Francisco Bay Area (mostly on-site). The robots are here.
• Duration: ~6 months, with flexible timings. Co-op and academic-term arrangements are welcome, with the possibility of a full-time offer based on performance.
• Start: As soon as possible (rolling applications)
• Scope: Personalized for each candidate. High end-to-end ownership of a core project, with a lot of breadth.