Internal deploy tooling
A small CLI that took deployment from a fifteen-step runbook to one command, and made rollbacks something people were willing to do.
- Year
- 2022
- Role
- Engineer
- Stack
- TypeScript, Docker, GitHub Actions
Placeholder content — replace with the real write-up.
The problem
Deploying meant following a wiki page that was subtly out of date. Rollbacks were rare because nobody was confident they’d work, which meant every bad release turned into a forward-fix under pressure.
What I built
A CLI wrapping the whole pipeline: build, verify, promote, and roll back. Every deploy recorded what it did, so the history was queryable rather than reconstructed from chat logs. Rollback became a first-class command tested on every release, not an emergency procedure.
Adoption came from making the tool faster than the manual path, not from mandating it.
What I’d do differently
I built configuration flexibility nobody asked for. Three services used the escape hatches; the other forty used defaults. I’d ship the opinionated version first.