NB: generalization should be one of, if not the first thing you think about and plan for in any ML project.