Summary
Qingnan Duan is a software engineer with 11 years of experience focused on accelerating large-model distributed training and production ML performance. He has driven performance and reliability improvements at Microsoft (Azure Document Intelligence) and now works on large-model distributed training acceleration at Google and DeepMind. His work blends low-level optimization (memory pooling, efficient IOU computations, RTree usage) with build and pipeline engineering to reduce latency and speed up CI/CD. Comfortable across cloud services and ML stacks, he has a track record of diagnosing worst-case performance and removing repeat computations to make systems predictable at scale. Based in North Hertfordshire, he combines rigorous Computer Science training from Tsinghua and Waterloo with hands-on production delivery in global teams. A pragmatic problem-solver, he often focuses on subtle efficiency wins that disproportionately improve end-to-end throughput.
11 years of coding experience