How do you as a DS feel about foundation models?

As an MLE focusing on MLOps who only does modeling for fun and not a 9-5, it’s our entire world right now. How do we scale up the fine tuning? What’s the best way to implement RAG? Which framework seems most legitimate for our use case?

But from a pure DS perspective what is your view?

case datascience foundation framework fun mle mlops modeling rag scale world