One of my favorite Jordan Volz ML Blogs is his 3 generations of ML Platforms (link in the comments). It's a Continual classic.
I take his foundation but instead of using his engineering lens, I offer the marketing perspective: research/experiment, end-to-end, and MLOPs.
I find this derivation useful when talking to teams evaluating tools, as as we're often starting the conversation with capabilities available in their existing toolset and the capabilities they desire.
As an aside, as the cloud providers bundle more and more stuff into Sagemaker, Vertex, and Azure ML, they essentially transcend my buckets. Whether or not they succeed in their aims is an exercise left to the user. But I should be explicit about that - no one can classify the multitudes that is a Sagemaker/Vertex/AzureML
Watch video The Three Types of ML Platforms online without registration, duration hours minute second in high quality. This video was added by user Gus Cavanaugh 13 January 2023, don't forget to share it with your friends and acquaintances, it has been viewed on our site 177 once and liked it 6 people.