What is the best and correct way of hosting an endpoint running R code?

Question

I think it must be a relatively common use case to load a model and invoke an endpoint to call R\'s predict(object, newdata, ...) function. I wanted to do this wit