Software that touches the world.
Machine vision, autonomous inspection, and sensor systems, designed by the people who built the robots.
See the workThe gap
Most firms that can write your app cannot instrument your equipment.
Most integrators who can instrument your equipment cannot write your app.
So the model comes from one vendor, the sensor from another, the controller from a third, and the dashboard from a fourth. When it does not work in the field, each one can point at the others and be technically correct.
We do all four. That is the whole argument.
What we build
Machine vision
Defect detection, grading, and counting on production lines and in the field. Trained on your material, evaluated against a held out set you agree to in advance, with the failure modes written down rather than hidden.
Autonomous inspection
Drone and ground platforms for corridors, yards, fields, and facilities. Route, capture, detect, and report as one workflow, not four tools.
Sensing and instrumentation
Sensor selection, mounting, calibration, and telemetry for equipment that was not designed to be measured.
Control and integration
The software between the model and the machine, including the parts nobody demos: power budgets, thermal limits, intermittent links, and what happens when a sensor lies.
Where the answer has to be local, we build the hardware it runs on. Cyberdecks →
The last mile
The hard part of physical AI is almost never the model.
It is power. Connectivity. Heat. Dust. Gloves. Daylight on a screen. A crew that has ninety seconds to look at a result before the next task. A site where the network ends at the gate.
We design for those first, because a model that only works on a bench is a demo, not a system.
Our applied lines
Agrobotics
Autonomous systems for growing, sensing, and screening. Precision phenotyping, canopy and trait measurement, and drone delivered intervention.
Drone technology
Airframes, payloads, autonomy, and the ground software, built as one system rather than assembled from a catalog.
Durable medical devices
Devices designed under design controls, with the sensing and embedded software built by the same team.
Where we have done this
Agriculture, drone agrobotics, and precision phenotyping, with published research behind the methods rather than a vendor datasheet.
See the research →How an engagement runs
| 1 | Site visit and scoping. We look at the actual equipment and the actual conditions. Remote scoping of field work does not work. |
| 2 | Fixed fee proposal. Scope, deliverables, acceptance criteria, and price, all written before anything starts. |
| 3 | Paid pilot, if you want one. Measurable success criteria agreed in advance. If it misses, you stop. |
| 4 | Build and field validation. Evaluated on your material, in your conditions, not on a public dataset. |
| 5 | Handover. Source, models, design files, documentation, and a walkthrough. IP transfers on final payment. |