> I'm having a really hard time figuring out what Watson is.

That is consistent with danpalmer's suggestion that it is a marketing construct.

I think sklearn/pandas are very low level tools that one could use to build products like those offered by the Watson APIs. Will Watson be better than a developer playing with sklearn for a week? Yes. Will it be better than a small machine-learning focused team, working on a domain specific problem for a few months? I would doubt it.

A fair comparison might be something like microsoft's cognitive services or google's cloud vision.