Over time, I've found different uses for it, and the most recent way that I'm using it is likely the one I'll settle on.

Authored by embedded ML specialists with extensive experience in ESP32 voice recognition architecture, TinyML optimisation, ...