AutoGen architect and Microsoft Agent Framework principal architect.
Shaping what agents on the Microsoft stack look like in production.
The architect behind AutoGen and the Microsoft Agent Framework - the person whose design decisions become the abstractions everyone else builds against.
Eric Zhu is the architect of AutoGen and a principal architect of the Microsoft Agent Framework. His work sits at the foundation of Microsoft's agent tooling: the core abstractions, the programming model, and the design choices that decide what building an agent on this stack actually feels like.
Those decisions matter far beyond a single project. When a framework defines how agents are composed, how they pass messages, and how tools and memory plug in, it shapes the mental model of every developer who adopts it - and Eric's architecture is what a large and growing number of teams are now building production systems on top of.
He comes to the work with a research background in data systems and is also known in the open-source world as the author of datasketch, a widely used Python library for probabilistic data structures like MinHash and LSH.
The frameworks and libraries the community most associates with him.
The multi-agent framework whose core abstractions and programming model shaped how a generation of agent systems get built.
Microsoft's open-source framework for agentic applications - the abstractions a large and growing number of teams now build production systems on.
A widely used Python library for probabilistic data structures like MinHash and LSH - his open-source work from a data-systems research background.
The work most associated with his name, and the themes that run through it.
The work he is best known for, plus where he shares the thinking behind it.