The problem with most guides and even ML frameworks (e.g. MLFlow) is that they do everything pretty well up to deployment, and then offer only very basic options that are not really fit-for-purpose for intermediate or advanced systems. It's definitely the biggest differentiator between the best resources and everything else