Machine Learning for RF Analog-to-Digital Signal Chain Autonomous Machines

Dear all,
I’m interested in receiving and transmitting RF signals. I would like to use a machine learning algorithm for waveform generation. Pynq looks like a very interesting option to easily integrate algorithms in python with the RF signal acquisition chain of the RFSoC boards.

Unfortunately, Pynq can only take care of the inference, requiring something external for training the models. Do you have any experience in integration/connection of Pynq boards with, for example, the Nvidia Xavier, or other similar platforms?