Capabilities
Physical AI isn't one skill — it's a stack, from embedded foundations to dependable production. These are the layers we engineer, each readable at executive, architecture and engineering depth.
Trust, Safety & Cybersecurity
Run, observe and recover a fleet across edge and cloud.
vLLM · TensorRT · fleet ops
Vision-language-action models, open-weight intelligence and industrial knowledge.
Mistral · VLM/VLA · industrial RAG
Seeing, mapping, planning and moving in the physical world.
ROS 2 · SLAM · Isaac Sim
The kernel-to-fleet software foundation that makes a device dependable.
Linux · Yocto · PREEMPT_RT · OTA
Sensors, actuators and deterministic control at the physical edge.
Sensors · MCU/PLC · OPC-UA · CAN
The full Physical AI stack — embedded foundations at the base, building up through perception, intelligence and operations, with a trust, safety and cybersecurity envelope around every layer.
The perception, control, and autonomy patterns we work in — illustrated across open robotics platforms.
Perception and human interaction on open platforms such as Reachy Mini.
Low-cost manipulation and imitation learning in the LeRobot / SO-101 ecosystem.
Aerial perception and autonomy on open hardware and software stacks.
Inspection and manipulation cells read by a perception layer.
Energy and readiness monitoring across an autonomous fleet.
Navigation and forward perception for mobile ground platforms.
Not sure what hardware a workload needs? Get the smallest role that fits your envelope.