We're building a physics-grounded, self-play-trained world model for general-purpose robot fleets. But the fastest way to know which use cases matter, where the hard problems actually are, and what "good enough" looks like isn't a whiteboard — it's a robot doing a real job at a real customer site. You'd own that: getting humanoid robots deployed on simple, well-scoped tasks using today's best available open-source software, and turning that experience into the customer insight and the baseline our research has to outperform.
_ Scope, install, and run humanoid robot deployments at customer sites for narrowly defined, high-value tasks — starting simple, expanding as it proves out.
_ Build on today's open-source robot software and control stacks rather than waiting for our own models to be ready; ship the best available solution now.
_ Work directly with customers across segments to find the strongest entry use cases and build real operational understanding of what they actually need.
_ Turn each deployment into a structured benchmark — a concrete bar that our internally developed models need to beat before they replace the open-source baseline in the field.
_ Feed what you learn (task structure, failure modes, customer constraints) directly back into the research team's priorities.
_ Hands-on experience deploying robots (humanoid or otherwise) on real tasks outside a lab — you've made something work in an environment you didn't fully control.
_ Strong working knowledge of open-source robotics software and control stacks, and comfort integrating and adapting them quickly rather than building from scratch.
_ Genuine interest in the customer side of robotics: scoping a use case, managing expectations, understanding what "reliable enough to trust" means for a specific operation.
_ Comfortable being the only person on the ground for a deployment — this is early-stage, hands-on, and often unglamorous work.
_ Prior experience in a customer-facing or field deployment role (industrial automation, applied robotics, systems integration).
_ Familiarity with benchmarking methodology — designing evaluations that are fair, reproducible, and actually predictive of real-world performance.
_ Background with humanoid platforms specifically, or general-purpose (non-task-specific) robots.
_ Comfort traveling to customer sites as needed.
