SABRE CORPORATIONCycle time: −88%

Sabre's airline shopping engine — the system that quotes flights when you search for a fare — was deploying quarterly. Mean time to detect a regression was three weeks. Mean time to roll back was measured in pages, not seconds.

Leadership wanted same-day releases without losing the four-nines uptime SLA they'd promised airlines. The platform team had two strong engineers; what they needed was an opinionated pattern and a 12-week head start.

GitOps with Argo CD on a multi-region Kubernetes cluster. Every change is a pull request against a manifest repo. Sync happens automatically; rollback is a revert.

Progressive delivery via Argo Rollouts — 1% → 10% → 50% → 100% with automated metric gates. If P99 latency or error rate breaches the SLO, the rollout reverses itself.

Replaced the legacy CI with a pipeline that builds, tests, signs, and pushes images in under 8 minutes. Trunk-based, with feature flags handling unfinished work.

Week 1–2: discovery. Read code, sat with on-call, mapped existing dependencies. Wrote a 14-page report and asked Sabre to push back. They did.

Week 3–10: build. Pair-programmed with the Sabre platform team — we were not a black box. Every PR carried two reviewers, one from each side.

Week 11–16: cutover. Two services per week migrated, each with a one-week soak period before the next. No big-bang launch.

Three years on, the platform Sabre's team operates without us still runs the same pattern. Deploy cadence: 14 production deploys on the busiest week we've checked. Sev-1 incidents related to deploy: zero in the last 18 months.

We still get a Slack message from the platform lead when a hairy migration goes clean. It's the highest-trust signal in this industry.