Please, help to find materials/additional information which explain how to use Human in the Loop Sagemaker API in production. What I mean is that in the Lab #3 we have an example where we send data to the model manually and then initiate Human in the Loop together with the piece of code that generates model result. But in practice when we deploy our models to the endpoint (like a separate API service) do we need to integrate that code (that triggers Human in the Loop) to the endpoint? Or that should be implemented into the program that send data to the endpoint, fetch the result and then start Human in the Loops? Should it be some kind of middleware or just a part of an application that implements business logic?
And another question what is the best practice to use Human in the loop Sagemaker API if I’m working with batch inference or if I want to collect a bunch of examples before sending them to the group of experts / volunteers for labelling?