About Hobson

I'm a solutions architect based in Shanghai. Over fifteen years I've gone from Oracle DBA to product and delivery for a distributed database, and on into financial services — securities, futures, insurance — working on database selection, PoCs, migration, disaster recovery and replacing foreign stacks with domestic ones. The most important skill along the way hasn't been memorizing product specs. It's been reducing a complex problem to something people can decide on, execute and verify.

Now I'm applying the same skill to enterprise AI. I focus on AI agents, MCP, RAG, agent memory, context engineering, harnesses and evaluation — and just as much on identity, permissions, sandboxing, auditing and observability. A demo that can answer questions is not the same as something ready for production.

My view of enterprise AI transformation: first find the high-value activities and dig out the know-how buried in the organization, then rebuild Human + AI workflows with Context → Decision → Action as the smallest unit. The goal is to go all the way from Know-how → Workflow → Agent → Capability, turning individual experience into organizational capability that can be reused, governed and keeps improving.

Outside work, health comes first, then reading, running and writing things down. I've kept a weekly journal since 2022. When I read, I'm not collecting quotes — I want to keep the judgments I'll be able to use next time. And I don't write to produce content; I write so my thinking leaves a trace.

Lately I've been building life-twin: turning years of notes into a digital twin with hybrid search, long-term memory and my own writing style. I also keep testing a new way of working with Codex, Claude Code, Dify, n8n and similar tools — people own the goals, the standards and the judgment, and agents execute within clear boundaries.

If you'd like to talk about enterprise AI, agent engineering or databases — or just swap reading lists — I'd love to hear from you.