Tell me about a time you had to make a trade-off between model accuracy and system latency. How did you make the decision?
