Can someone explain me , what is the USP of Ray over Sagemaker/VertexAI, beyond being a cloud agnostic solution. Just trying to understand , so that i can make a case to my management
The TLDR is that in the realm for specifically distributed computing for machine learning applications it’s faster and cheaper because of a whole stack of clever optimizations that Ray does for you.
This is particularly valuable in industry today because everyone has to increasingly turn to distributed computing for solving problems as just scaling up is no longer feasible for a variety of reasons in the AI/ML space.
This includes scenarios data ingest, training, tuning, offline batch inference, and online serving.
There are a ton of examples of how companies are finding success with this: see https://www.anyscale.com/blog.
You can also cross reference to those companies own blogs to ensure a second opinion/fair evaluation.