Skip to content
Cheng-Han Lin

About

I’m Cheng-Han Lin, a full-stack developer based in Taichung, Taiwan.

I graduated with a B.S. in Computer Science and Information Engineering from National Chin-Yi University of Technology in June 2026, and started the M.S. programme at the same university that September.

What I work on

My work keeps returning to one problem: how do you turn a large language model into something people can trust?

The model itself is the easy part to obtain. The hard part is everything around it — making sure it cannot invent an answer, keeping sensitive data inside a controlled environment, producing a useful forecast from only a few hundred rows per user, and deciding how the system should behave when it is wrong.

Concretely, I have built:

What I care about

I care less about which technology was used than about why it was chosen.

Every decision costs something. Choosing Random Forest gives up whatever temporal structure a sequence model might have found. Tuning for recall means accepting more false alarms. Self-hosting a model means owning its operations. The difference between engineers shows up mostly in whether they can articulate those trade-offs.

The HeartBox case study on this site is written to that standard.

Contact