jarm.ai is a humanoid robotics company building both the robot and the foundational software required to scale physical AI.



Humanoid robotics is entering a phase where rapid progress is possible but only if systems are designed to learn quickly, validate assumptions early, and remain extensible as complexity grows.



We develop humanoid hardware alongside jarmOS, a unified control and execution stack that connects perception, language, tools, and agents to reliable physical action. The interface between high-level intent and execution is a core focus, because it determines how well humanoid systems adapt as models, hardware, and environments evolve.



Rather than optimizing early for narrow use cases or attempting to build a fully general humanoid all at once, we sequence development to maximize learning speed while keeping scale in sight. An upper-body humanoid platform allows us to test real-world constraints, stress integrations, and iterate rapidly, while building toward full-body humanoid systems.



Our approach enables fast internal progress and external experimentation, accelerating validation beyond what a small team could achieve alone. This creates a path to scale that is grounded in real usage, not assumptions.



Based in Europe, jarm.ai works with research and technical partners while building toward deployable humanoid robots.



We believe scalable humanoid robots emerge from architectures that are validated early and designed to grow.