In most of the cases, the predictions are meant to be delivered via a REST API. One of the approaches is to pre-compute predictions -> store in DB -> serve via API.
Do you have a better pipeline for your production models?
No comments yet.