How would you architect a system to serve low-latency AI predictions to a global user base at Altimate.Ai?
