Robots cannot learn work they have never seen. Metari builds and operates the environments where that work happens, then holds them still so progress can be measured. Collect, train, deploy.
A short film illustrating the concept end to end: the campus, expert data capture, evaluation, experience, and the path to deployment at scale.
Experienced professionals do the real job at working speed, in rooms built for capture. Worn rigs and fixed ceiling, wall and working-height cameras record it on one clock, while the operator explains the decision behind each move and the standard it has to meet.
Run the policy on the same task, in the same room, against the same acceptance criteria as the human demonstration. Nothing between the two runs has moved, so a change in the score is a change in the model and not a change in the building.
The building that produced the data becomes the floor where operators watch the work before they buy, then one of the first properties the robot actually works in, with the team that trained it already on site.
Every wearable programme running inside occupied buildings is reactive: it records whatever happened to occur that day. Occupancy sets the ceiling, guests are present, and a good day yields one or two usable room turns.
A dedicated environment inverts that. The partner names the scenario and we produce it, on schedule, to volume. No guests, reset and re-run on a shift pattern, against a task list the partner writes. Clutter, lighting, linen, spills and operator are changed deliberately between runs. Spills, breakages and maintenance faults can be staged on purpose, which is the long tail a live-building programme will almost never catch on camera.
Our campus concept brings public demonstrations, controlled task environments, and technical teams together in one setting.
Concept visualization
Explore humanoids in an elegant public gallery, with supervised training activity visible through glass.
Click through the environments. Every image is a concept visualization.
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Concept visualization
Illustrative concept
Metari’s vision is to build the infrastructure that helps physical AI move from development into everyday operations.
Its campuses would combine expert data collection, structured evaluation, and elevated experiences where enterprise buyers and the community can explore humanoid capabilities through guided demonstrations and supervised participation. Planned initial revenue would come from training data and evaluation services, with deployment support and ongoing service expanding over time.
The network could grow through leased spaces, converted properties, and property investment partners, while Metari develops the software, operating processes, and customer relationships connecting each location. The long-term advantage would come from accumulated expertise, useful data, repeatable evaluations, and deployment feedback.
A description of Metari’s concept and direction, not of existing operations.
The campus is where operators and the public can see humanoids work in real settings, and where robots are prepared, evaluated, and readied for the world. Experience builds belief; belief drives adoption.
Concept visualizationA recording shows what happened. Expert context helps explain why it happened, whether it met the standard, and how to recover when it did not.
Illustrative example
An illustrative interface. Not live sensor data.
A successful demonstration is a starting point. Evaluation asks what happens with a different layout, unfamiliar objects, or a need for human help.
Illustrative evaluation · task: place folded towels on a shelf
An expert demonstrates the task in a known layout, establishing the standard for a correct result.
The shelf position and arrangement change, and the objects are presented differently.
Relationships with experienced operators and professionals who know how tasks should be performed.
Configurable spaces that introduce new layouts, objects, conditions, and exceptions.
Task definitions and evaluation methods that make observations comparable across runs.
Lessons from pilots that guide what to demonstrate, test, and improve next.
Collection that stops at delivery is a one-off purchase. The same rooms, held still and run as an evaluation environment, are what turn a data sale into a standing programme. Every pass returns a targeted brief, not more of the same.
Our intended advantage is the combination: expert relationships, useful data collected with appropriate rights, repeatable evaluations, and experience that improves the next engagement.
The near-term business is expert data and model training. The campus network and deployment layer scale on top of it.
Expert demonstration data and model training, sold to robotics and physical-AI teams. The revenue that exists now.
Recurring evaluation of humanoids against real-world task standards as their models iterate.
Readiness, deployment support, and the campus and real-estate layer that follows adoption.
Hospitality gives us ready-made environments. The concept is to grow the campus network through real estate, not build every site from scratch.
Work with operators to instrument and run blocks or floors of existing hotels as live capture and demonstration space.
Buy existing hotels and convert them into Physical Intelligence Campuses: experience, capture, and deployment under one roof.
Partner with REITs and funds that own the land and buildings while Metari operates and converts them, and acquire select properties directly.
The properties within reach hold room inventory in the thousands, across multiple markets and property classes. Capacity is added by taking floors out of inventory and instrumenting them, not by constructing buildings. The camera rig and its calibration are the expensive part, and they never move.
Each property becomes a compounding asset: an experience destination, a data-generation engine, and a deployment hub. A concept for how the network could scale, not a description of existing operations.
A closer look at how experts, instrumented environments, and evaluation could come together on a single campus.
Discuss expert demonstrations, task variation, dataset requirements, and evaluation design.
Discuss a data or evaluation projectExplore how your operating expertise and environments could contribute to future data collection and supervised pilots.
Explore an operating partnershipLearn about the campus concept and the infrastructure model behind it.
Connect with the founderHelping humanoids learn how real work should be performed and evaluated, from someone who has spent a career defining and running it.
Steven Ortmann brings more than 20 years of hospitality operating experience, including leading teams, defining operating procedures, and managing service quality. Metari applies that perspective to a new challenge: helping humanoids learn how real work should be performed and evaluated.
Tell us about the task, environment, or partnership you have in mind.