AI Interviewer Moving on to system design. How would you architect a rate limiter for a high-traffic distributed API?