"Edge" unfortunately is a broad term whose definition depends on the scope of the system. But "Edge ML" is the precise way to communicate what we're doing.

I agree that "on device ML" is less ambiguous, so will give this some more thought.

Thanks for the suggestion!