Interview question asked to Data Scientists interviewing at Monzo, Yahoo, Scribd and other companies. Original question asked: Describe how you solve the "starvation vs. latency" dilemma when implementing inference batching on constrained compute..