I've used a track splitting tool from Spleeter, explored music creation GANs and looked into automated audio recognition using Tensorflow