ZhangDonghao

Career Goal: AI-HPC Domain Expert

ZhangDonghao is an AI infrastructure engineer focused on the convergence of AI and High-Performance Computing (HPC). He has long worked on building and delivering intelligent computing clusters, high-speed networks, AI platforms, and scheduling systems, leading multiple large AI/HPC cluster projects, and continues to run the AI-HPC.org technical portal and its supporting toolchain.

Work Experience

Operations & Development Engineer (AI Infra)

• Intelligent computing cluster infrastructure support: Provided technical support for a carrier-grade 10K-GPU AI training cluster in Shanghai, responsible for InfiniBand high-speed network architecture optimization, performance monitoring, and fault diagnosis to ensure communication efficiency for distributed training tasks.
• AI platform & toolchain development: Led the development of an AI Agent service platform with Python (Flask/Django), encapsulating underlying compute APIs; used Docker containerization to rapidly deploy and manage AI development and inference environments (Dify, ComfyUI, etc.), improving R&D efficiency for algorithm engineers.
• Automation & monitoring: Wrote Shell scripts and developed automated alerting and log analysis tools for real-time monitoring and rapid response on cluster key metrics.
• Cloud-native & DevOps: Managed the Alibaba Cloud Yunxiao DevOps platform, built CI/CD pipelines, and performed cost forecasting and performance evaluation for cloud resources of applications such as intelligent customer service.

AI & HPC Product Delivery Manager (AI Infra)

• HPC cluster construction: Led the delivery of multiple large AI/HPC projects, including the Tsinghua University GPU cluster and Peking University HPC projects, covering the full lifecycle from hardware racking and cabling (servers, storage, IB switches) to OS, scheduler, and AI software stack deployment and acceptance.
• Heterogeneous compute resource management: Deeply proficient in Slurm, responsible for unified CPU+GPU heterogeneous scheduling and management strategy, writing scheduling scripts to optimize resource allocation and execution efficiency for scientific computing and AI training workloads.
• Performance testing & tuning: Led comprehensive performance benchmarking (HPL, NCCL Tests, etc.) on CPU, memory, distributed storage, IB network, and GPU, analyzing bottlenecks and performing system-level tuning to ensure optimal cluster performance.
• AI platform deployment & maintenance: Core lead for deploying and operating the Inspur AIStation AI platform, providing one-stop model training, resource monitoring, and task management services, plus technical training and remote support.

IT Technical Support Instructor

• Taught Linux, Shell, and Python courses and guided students in deploying distributed and containerized software such as Docker, Nginx, and Ceph.
• Built and maintained teaching lab environments with Prometheus + Grafana for server cluster monitoring.

Skills

• Programming: Proficient in Python and Shell; familiar with C/C++ with basic CUDA programming.
• Distributed technologies: Familiar with HPC cluster architecture and InfiniBand networks; proficient in Docker containerization; familiar with Kubernetes (K8s) orchestration.
• AI frameworks & tools: Familiar with PyTorch/TensorFlow training workflows; understands fundamentals of Transformer and CNN models.
• Scheduling & management: Proficient in Slurm; familiar with heterogeneous (CPU+GPU) task scheduling.
• Monitoring & operations: Skilled in Prometheus and Grafana; experienced in Linux system administration and performance tuning.

Certifications

• NVIDIA: LLM applications with prompt engineering, accelerating C/C++ applications with CUDA, and developing DPU applications with DOCA

Selected Works

• AI-HPC.org — this portal: an AI-HPC expert site on AI computing × HPC, 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
• 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