ZhangDonghao
Career Goal: AI-HPC Domain Expert
ZhangDonghao is an engineer focused on the convergence of AI and High-Performance Computing (HPC), with long-term experience in the engineering implementation of AI infrastructure. He runs AI-HPC.org and its supporting technical platforms, dedicated to turning the knowledge of AI clusters, accelerators, compilers, and large-model engineering into reproducible engineering practices.
Expertise
- AI Infrastructure: AI cluster architecture, high-performance networking (IB/RoCE), parallel storage, scheduling & cloud-native (Docker/K8s/Slurm)
- AI Accelerators: MLU accelerator driver, operations, and fault diagnosis; OAM/UBB ecosystem
- AI Compilers & Frameworks: PyTorch internals, Triton, MLIR/TVM compiler technologies
- Large Model Engineering: distributed training (3D parallelism), inference optimization (vLLM), RAG/Agent applications
- SRE / DevOps: automated provisioning, monitoring & alerting, log analysis, and toolchain building
Selected Works
- AI-HPC.org — this portal: a knowledge alliance platform for AI & scientific computing convergence, with a 14-chapter AI/HPC engineering knowledge base
- AI-HPC Technical Expert (chat.ai-hpc.org) — an AI-HPC Q&A and navigation assistant built with Dify
- MLU Log Analysis Bot — Feishu/WeCom bot: upload a log bundle to get automatic MLU fault diagnosis, card-replacement decisions, and ledger records (log-analyzer / cnlogspector)
- AIInfra Knowledge Base — systematic AI infrastructure knowledge (cluster / storage & network / cloud-native / training / inference)
- Technical Blog Series — deep dives into PyTorch internals, Autograd, Triton/GPU programming, NCCL communication, and compiler frameworks
Contact
- Email: openaihpc@gmail.com
- GitHub: github.com/openaihpc