NVIDIA 如何用 AI 代理简化 AI 工厂运维
原标题：How NVIDIA Uses AI Agents to Streamline AI Factory Operations
AI 摘要
NVIDIA 发布技术博客，介绍其内部开发的 AI Factory Operations Agent，用于简化 NVIDIA DGX SuperPOD 的运维工作，减少站点可靠性工程师（SRE）的人工排查负担。该框架基于 NVIDIA NemoClaw 蓝图和 OpenShell 运行时构建，多个子代理通过 MCP 协议通信，分别负责作业恢复、硬件故障、可观测性、研究、Kubernetes 和集群管理等领域，并通过聊天界面提供根因分析和恢复建议。系统采用分层安全设计，包括只读 kubectl 工具、RBAC 和每会话 Kubernetes 沙箱，并支持本地部署与数据主权。
正文节选
As GPU clusters scale, operations span compute, network, hardware, and scheduling. Site reliability engineers (SREs) often have to examine multiple dashboards to identify the cause of training performance drops. What if a team of AI agents could perform that correlation through one interface? NVIDIA developed the AI Factory Operations Agent to streamline its own internal NVIDIA DGX SuperPOD operations and reduce the manual effort by NVIDIA SREs. Sub-agents communicate through Model Context Proto