I should probably learn how inference is served at scale. Any have a zero to one reference for this?