[Client] builds a robot that moves goods around a warehouse. The mechanics existed; the robot had to see where it was, plan a route and get there without a human on a joystick. The team that built the hardware had nobody to write the navigation and the software that manages a fleet of them.
What we shipped
Track detection from the onboard camera with OpenCV and an ONNX model, running on the robot itself.
Map plotting from what the robot sees, and A* pathfinding over that map.
Navigation code deployed to the robot's [Raspberry Pi / controller] on linux-arm64.
A web back-office to send jobs, watch position and state, and review logs.
What it changed
[Outcome with a number: robots in operation, hours per day, error rate, pick throughput, or "running since 20XX".]
[Photo or short clip frame of the robot]
Tech: .NET, OpenCV, ONNX, Raspberry Pi (linux-arm64), Blazor back-office, Azure.
SLACrew