Is there a trick to deploying audio classification models to this board?
As the XIAO BLE Sense is not an officially supported board, you will need to write your own inference code. Take a look at the Nano 33 BLE Sense continuous example in the Arduino library–that should show you how to perform continuous sampling with the microphone with windowing (e.g. using a ring or double buffer) to record slices of audio to send to the EI library for inference. You will need to change the microphone sampling code in that demo to work with the microphone library for the XIAO nRF52840, as the microphones (and libraries) are different.