If a portfolio answers “what have I built?”, this timeline tries to answer a deeper question: why I became Xiaohui, and how my judgment was upgraded at each stage.
I have long cared about three kinds of questions:
- How to make complex systems reliable.
- How to bring AI Agents into real business workflows.
- How to keep learning, expressing, and creating in a fast-changing environment.
Earlier Years: Turning Curiosity Into Long-Term Ability
Peking University left me with more than specific knowledge. It trained three habits: keep questioning, build judgment under uncertainty, and organize knowledge through writing while validating ideas through projects.
Those habits later became part of my technical blog, open-source work, and Agent practice.
2021: From Student to Engineer
After graduating in 2021, I joined Alibaba and began learning engineering through real business systems. Code is not finished when it compiles. It must survive complex chains, collaboration, and long-term evolution.
This period shaped my preference for high-quality development: testing, maintainability, clear boundaries, and modeling business problems properly.
2022: Rebuilding Business Systems
I became more deeply involved in business system construction. The hard parts were not isolated technologies, but changing requirements, data consistency, link stability, business semantics, and coordination cost.
I gradually realized that an engineer’s core ability is not using many tools, but giving reliable solutions under complex constraints.
2023: Full Stack and Business-Finance Integration
I worked on Taogongchang finance, billing, invoicing, and business-finance integration. The closer a system gets to business substance, the more it needs technical abstraction and business semantics to be considered together.
Engineering is not only implementing requirements. It also means understanding the business loop behind funds, accounts, and invoices, identifying system boundaries, and keeping solutions deliverable across roles.
2024: From Engineering to AI Agents
I moved from traditional business systems toward AI Agent delivery. I worked on the Taogongchang customer service knowledge base and later on Ant Insurance risk-control case analysis driven by AI Agents.
The biggest shift was that I began to treat Agents as a new software architecture paradigm, not merely a smarter chat entry point.
2025: From Demo to Reliable Agent Systems
I entered the deeper water of Agent engineering: MCP, context engineering, Agent runtime, tool ecosystems, and sustainable delivery.
Key experiences include contributing to AgentUniverse, building MCP-related capabilities for Ant Insurance, creating MCPAdvisor, and leading hackathon projects at OceanBase and Ant Group.
2026 and Beyond
My current focus is to turn Agent practice into reusable engineering systems: clearer protocols, better context management, reliable runtime behavior, and knowledge that can travel beyond a single project or company.