An IP69 Box for Jetson-Class Vision

Syslogic is building rugged computer systems around NVIDIA's Jetson Orin Nano 2 for computer vision and inference — person and object detection in mobile machinery, autonomous vehicles, robotics and intelligent vision systems. Versions are rated IP67 and IP69 and specified from minus 25 to plus 70 degrees Celsius. Product manager Michael Jung says the compact AI Rugged Computer RPC RS with Jetson Orin Nano 2 launches in the first half of 2027.

The interesting engineering is not the module, which anyone can buy, but the enclosure specification. IP69 means the housing withstands high-pressure, high-temperature washdown — the test involves water at 80 degrees delivered at 80 to 100 bar. That is not a general ruggedness claim; it is the specific requirement of equipment that gets cleaned aggressively, which covers food and beverage processing, agricultural machinery and anything working in mud or slurry.

Meeting it while dissipating the heat of an inference accelerator is the real problem. A sealed enclosure cannot use a fan or a vent, so all heat leaves by conduction through the housing, and the thermal design becomes the constraint on how much compute the box can carry. That is why rugged AI hardware trails the desktop equivalent by a generation or more, and why the module choice — Orin Nano rather than a larger Orin — is a thermal decision as much as a cost one.

The temperature range matters for a subtler reason. Minus 25 to plus 70 is not about survival but about specified operation, which means the silicon must hold its clocks and the storage must remain writable across that range. Industrial-grade components with that specification cost more and are frequently the reason a rugged product exists at all, since the same computation runs perfectly well on a commercial board in a comfortable room.

The application list points at where machine vision is actually growing. Person detection on mobile machinery is a safety function — a loader, a harvester or an AMR that can see a person in its path — and safety functions bring their own requirements around determinism, diagnostics and failure behaviour that go well beyond running a model. The 2027 launch date puts the product in the same window as the fleet of construction and agricultural machines now being specified with perception as standard rather than as an option.