
Protect your model. Wherever it runs.
Your classifier ships inside an OEM machine or onto a customer's line. It stays encrypted there: no plaintext in memory, nothing to copy off the industrial PC.
Where the model walks away today
01The OEM deal abroad
The partner in Asia or the US asks for escrow or perpetual runtime rights to your model.
02The customer who wants to self-host
The end customer wants your model on its own servers, where its staff can reach the file and the memory.
03The cloned machine
A copyist can rebuild the hardware. Your trained classifier is the one asset they cannot build, unless they can lift it.
Where it runs
- Jetson and GPU industrial PCsDocker container on ARM or x86
- Industrial edge platformsMapped in the architecture review
- Vision pipelinesONNX or TensorRT model in
- CadencePer batch and per event
Fit with your specific vision pipeline, including Siemens Industrial Edge, Picsellia or Robovision setups, is mapped in the architecture review.
In scope: recipe, validation, batch and diagnostics layers. Out of scope: belt-speed inline inference.
Keep your dongle
CodeMeter or Sentinel
Entitlement: who may run the model, where and for how long.
Wodan
Confidentiality of the model during inference.
They work side by side.
Pilot
- Your model converted and running encrypted on your target hardware
- A benchmark report: latency, accuracy delta and memory, on your hardware
- A deployment recipe
Bring one model and one deployment.
We map which layers fit your hardware and risks, and what a paid pilot would confirm.
Our engineers review your model and deployment context with you.
