Your hardware

The guides in this section walk you through compiling, profiling, and running inference on your own devices — a Raspberry Pi, NVIDIA Jetson board, desktop GPU, or any Linux machine reachable over SSH. This is ideal when you want:

• Fast feedback loops — jobs run directly on the target device. With ONNX Runtime, a full compile-and-profile cycle for a MobileNetV2 finishes in under 20 seconds — compared to ~10 minutes for the same model on a cloud provider.
• Full control — use any device you have, not just those available in a managed cloud.
• No cloud dependency — all computation stays on your network.

If you need to test across many different edge devices without purchasing and managing them, see the Cloud guides instead — cloud providers give you access to a wide range of hardware from your laptop.

Prerequisites

Before following any guide in this section, make sure you have completed the setup guide to:

• Create an Embedl Hub account
• Install the embedl-hub Python library
• Configure an API key

SSH access

Your device must be reachable over SSH with passwordless authentication. If that’s not set up yet:

1. Generate an SSH key pair (if you haven’t already):

   ssh-keygen -t ed25519

2. Copy the public key to the target device:

   ssh-copy-id user@192.168.1.42

3. Verify that you can connect without a password:

   ssh user@192.168.1.42 echo "Connected!"

Available guides

• ONNX Runtime — compile, profile, and invoke ONNX Runtime models on any Linux device using embedl-onnxruntime.
• TensorRT — compile, profile, and invoke TensorRT models on NVIDIA GPU-equipped devices using trtexec.