Patronus

Tell us what you are building.
We reply in person.

For a Runtime Engine request the form asks about models, devices and targets, so the first call can get specific straight away.

Your request

Key facts

Rough answers are fine. Whatever is still open, we sort out in the call.

Which models do you run or plan to run? (optional)
Where do the models run?

What happens next.

For a Runtime Engine request:

  1. Send the key facts

    Which models you run, on how many devices or instances, on which CPUs, and what you need for latency and memory. The form asks exactly that.

  2. Technical call

    We reply in person by email and suggest times. In about 30 minutes you walk through your setup with the people who build the models and the runtime, and agree on what to measure.

  3. Measurement on your hardware

    We measure ONNX Runtime against the Patronus Runtime Engine with your models and typical input lengths on your CPU family. You get latency and memory side by side.

  4. Offer and integration

    The written offer depends on the number of devices or instances and the term. Once it is signed, we help you integrate.

For every other topic we read your message ourselves and reply by email.

Reach us directly