Mingming Pei
Head of AI Platform Technology, NetEase Intelligent Enterprise
Responsible for the technical development and deployment of NetEase’s King Crab enterprise AI Agent platform, with a focus on AI compute resource scheduling, skill assetization, and multi-agent collaboration. The platform covers skill asset management, compute resource scheduling, and multi-agent collaborative governance, aiming to drive the evolution of AI Agents from standalone tools into organization-level collaborative platforms. He has also shared his experience in implementing cloud computing and AI platforms at numerous technology conferences in China and abroad.
Topic
From “Using Agents” to “Organizational Productivity”: Designing and Implementing Multi-Agent Systems with NetEase Enterprise’s “Emperor Crab”
As agents move from standalone assistants to real-world business collaboration, the challenge is no longer whether models are smart enough, but how to enable multiple agents to divide responsibilities without losing control, collaborate with traceability, deliver verifiable results, and continuously accumulate experience. This talk uses NetEase Enterprise’s “Emperor Crab” as a case study to introduce a multi-agent architecture designed for enterprise scenarios. A central agent is responsible for understanding goals, generating task graphs, coordinating dependencies, and consolidating results, while domain-specific sub-agents execute individual tasks as the sole owners of their respective responsibilities. Through Skills, contextual memory, permission governance, observability, and an evaluation loop, the system forms an operable, scalable platform for agent collaboration. The core idea is simple: Agents provide local intelligence, while the platform provides global order. The value of NetEase Enterprise’s “Emperor Crab” is not simply adding more agents, but transforming AI capabilities into reusable, governable, and continuously evolving organizational assets.
