Describe a time when you had to make a difficult trade-off between model accuracy and system latency. How did you decide?
