Engineering Deliveries & Certification

Chapter focus: Turning "things done" into "reusable methodology." A delivery retrospective is not a log — it breaks a large cluster delivery into architecture decisions → acceptance process → troubleshooting records, so the next project can reuse it directly; the certification guide turns expertise into a verifiable path.

Delivery Retrospectives

1. Tsinghua GPU Cluster Delivery Retrospective — Full-process breakdown of a large AI training cluster from architecture design to acceptance.
2. PKU HPC Project Delivery Retrospective — HPC cluster delivery for scientific computing: mixed workloads, the MPI ecosystem, and scheduling strategy.

Certification & Capability Path

3. NVIDIA Technical Certification Guide — The certification landscape, a preparation roadmap, and practical strategy for turning years of work into verifiable expert credibility.

How to Use

• Doing deliveries: read the Tsinghua retrospective for the process skeleton, then the PKU retrospective for the scientific-computing view; use both checklists at the end.
• Building expert credibility: retrospectives prove "I have done it," certifications prove "I can do it" — together they form a complete evidence chain.