Machine Learning Model Input and Output Flow

I am working on a backend for structuring and submitting data into a ML model. I have 3 questions regarding this process.

1. What is the best method to feed the model continuous data (updated every 30m interval)
2. What is the best method to deliver the datasets? (There are two that are being compared, should I consolidate the data on my side, or leave that to the model?)
3. How should the results be exported from the model and input into my database?

I am asking these questions as a software engineer seeking to bring a data scientist into the project so I can tell them what we can expect our workload to be.