How would you optimize a model for latency, accuracy, and computational cost simultaneously at Motorola Solutions?
