Mobile inference field guide

The model
in your pocket.

How to turn a trained model into a fast, private mobile feature—then prove the GPU is actually doing the work.