Walk me through a complex computer vision project you recently completed. What were the primary bottlenecks, and how did you resolve them?
