MingHsuan Lee
DATA SCIENTIST · ML ENGINEER · AUCKLAND, NZ
I write about AI agent infrastructure — MCP, context engineering, retrieval, and orchestration at production scale.
The notes here come from shipping multiple production AI agents on small, self-hosted infrastructure. The trade-offs, the decisions, the things worth knowing before you ship.
Latest writing
Archive →-
Infrastructure for solo AI builders: why I split one server into three
Running your own AI agents and GenAI tools? Here's the three-tier setup I landed on — dev, production, and a shared AI + data tier — and the three principles behind it.
Running AI in Production -
Embedding choice for production: when bge-m3 beats OpenAI
Why I run bge-m3 locally for daily workloads even though OpenAI's text-embedding-3-small is faster and cheaper at scale — a production decision framework.
Context Engineering -
minghsuanlee.net is Live
Get new posts by email
One email per post. Long-form notes on AI agent infrastructure. No promotions.