Hi! I came across the website a few weeks ago when looking for data to train a style transfer algorithm on. I was looking into timbre transfer, which is effectively a technique which transforms the sound of one instrument such to make it sound like a different one. This might be a cool video if you're interested in knowing more: https://youtu.be/YQAupr7JxNY
Such systems do require an insane amount of (good and isolated) data however and this website proved to be an awesome source for training data. Now I'm looking into a different music processing task for my master thesis, namely music source separation. That is the task of separating an input signal into two or more components (e.g. isolating the vocals from the instrumentals, good for karaoke!) and I think that we might be able to help each other out! I'll probably make a separate post on this in the forum.
In any case, thanks for having me here!