
You are choosing a model for a supervised learning system that will run at large scale in production. Several candidates are on the table, and the team needs a clear way to compare them before rollout.
What trade-offs do you consider when selecting a model for large-scale deployment?