Ask HN: How to Handle One Machine Learning Model by User?

Hello fellow data people, I am developing an application that I hope I could present you very soon based on Django. I would like to be able to compute one Machine Learning (ML) model for each of my users to recommand him ways to optimize his publications and auto-tagging.

It seems obvious that the algorithm should adapt itself to each of the user (How ? Retraining ?) and how do you manage a pool of ML models (store where ? how to monitor ?)

I guess this kind of thing is very similar to recommender systems adapted to each users or to spam filters which learns your preferences and so on.

Thank you in advance for your answers,

Best regards from France, Jonathan