Shadow Deployment

A deployment strategy where a new model runs alongside the current production model, receiving the same real-world inputs but without its outputs being shown to users. Shadow deployment lets teams evaluate a new model's behavior safely before committing to a switchover, surfacing failures and edge cases without exposing users to them.