How do you optimize latency for real-time inference in an edge-computing robotics scenario at Bosch Group?
