### ##### # # ### ##### ##### # # #####
# # # ## # # # # # # # #
##### # # # # ##### # # # # ####
# # # # ## # # # # # # #
# # ##### # # # # # ##### # #####
AI Native Systems Architect · Open-source Builder
I design and build the infrastructure, tools, and products that turn AI models into working systems.
My work spans enterprise AI governance, agent tooling, and the workspaces where people use AI every day.
| Project | Focus | What it enables |
|---|---|---|
| TokenHub | Enterprise AI governance | Govern model access, routing, permissions, usage attribution, budgets, and provider reconciliation. |
| Aer | AI-native browser workspace | Bring organized workspaces, page-aware AI, and agents into everyday browsing and work. |
| DeepChat | Local-first agent desktop | Work with models, tools, Skills, and agent runtimes in one desktop application. |
| Go-MCP | Agent tool connectivity | Connect AI applications with tools and external systems through a Go implementation of MCP. |
- Start with the work. Design around real tasks, context, tools, and the people using the system.
- Make control part of the architecture. Treat permissions, observability, reliability, and cost as design concerns from the beginning.
- Build for change. Use clear interfaces so models, tools, and workflows can evolve independently.
I bring the same focus on reusable infrastructure and developer experience from my work in Go to the AI systems I build today.
- Beego — Creator of the Go web framework for web applications, APIs, and backend services.
- Build Web Application with Golang — Author of the open-source book on building web applications with Go.
- Earlier work includes beedb and contributions to Swagger tooling for Go.
Interested in AI infrastructure, agent systems, and practical AI-native product design.
For project questions, start with the relevant repository's issues or discussions. For collaboration: xiemengjun@gmail.com.