Cloud
Compile, profile, and test models on managed device clouds.
Cloud providers let you compile, profile, and invoke models on real edge devices without owning any hardware. Jobs are dispatched to a managed device cloud and results are returned automatically. This is ideal when you want to:
- Test across a wide range of devices — compile and profile on Snapdragon-powered phones, automotive SoCs, and more, all from your laptop.
- Produce device-optimized models — cloud compilers apply INT8 quantization tuned for the target hardware.
- Share results with your team — every run is recorded on your Hub page for easy comparison and collaboration.
Cloud jobs are queued and typically take several minutes to complete — for example, compiling and profiling a MobileNetV2 on Qualcomm AI Hub takes around 10 minutes end to end. If you need faster feedback loops for experimentation, see the Run on Your Hardware section, where the same model completes in under 20 seconds with ONNX Runtime. The tradeoff is that cloud gives you instant access to dozens of edge devices you don’t have to purchase, configure, or maintain.
Prerequisites
Before following any cloud guide, make sure you have completed the setup guide to:
- Create an Embedl Hub account
- Install the
embedl-hubPython library - Configure an API key
Some cloud providers require additional setup — each guide covers the specifics.
Available guides
- Qualcomm AI Hub — compile, profile, and invoke TFLite and ONNX Runtime models on Qualcomm Snapdragon devices.
- Embedl Device Cloud — compile locally and profile TFLite models on the Embedl-managed device cloud backed by AWS Device Farm.