I build and operate the infrastructure behind large scientific facilities — particle accelerators, mostly, across national and European laboratories. The repositories here are the tools that work produced: build and deployment environments, IOC and console lifecycle utilities, provisioning playbooks, documentation pipelines, and mitigation scripts for legacy systems.

Most of them exist because something on the floor did not match what the records said it was.

• Program delivery — work breakdown structure, schedule and resource planning, commissioning playbooks, operational handover
• Physical infrastructure — rack layout and installation, cabling, network topology and bring-up, equipment receiving and asset tracking
• Control systems — EPICS-based distributed control, IOC lifecycle, console and terminal tooling
• Build and deployment — reproducible environments across Rocky Linux and Debian, Dockerized CI/CD, versioned releases
• Records and verification — single-source-of-truth documentation pipelines, auditable approval lifecycles, non-conformance reporting
• AI-assisted engineering — agent workflows for documentation, review, and provisioning, with skills, memory, and environment kept in one shared repository

Infrastructure fails where the digital record drifts away from physical reality. Not because anyone lied, but because a record can only hold what was measured. So I build the pieces that hold the two together: one dataset behind every drawing and configuration file, trails that record what was held back as well as what shipped, and verification that says how many out of how many.

I write about this at jeonghanlee.github.io/essay-site.

I may be slow to respond.