When you say you are not just doing inference, you mean you are also training your own llms? I am curious what other things can be done.

Fine tuning, and yeah training my own, experimenting with architectures and learning how it all works. Been a lot of fun