Kiln
Sable had forty engineers and one person who understood the deployment pipeline. We replaced tribal knowledge with a platform any engineer on the team can run.
- Challenge
- One person could deploy, once a week
- Outcome
- Any engineer, forty deploys a week
- Client
- Sable Technologies
- Services
- Cloud & DevOps + Custom Software
- Industry
- Software
- Year
- 2025
Kiln release dashboard mid-rollout
The challenge
Sable had forty engineers and one deploy pipeline that one person understood. Releases went out on Thursday evenings because that was when he was free.
Nothing was technically broken. The bus factor was one, the pipeline took ninety minutes, and nobody could tell you what was in a release without reading the commit log.
- Release duration
- 90m → 4m
- Engineers who can deploy
- 1 → 40
- Deploys
- 6/wk → 40/wk
- Rollbacks needed since
- 0
Our approach
We did not rewrite the pipeline. We made it legible: every step named, every step reversible, every step something you could run yourself and watch.
The platform is a thin layer over tools the team already had. That was deliberate — a bespoke deploy system is just a new single point of failure with our name on it.
“I went on holiday for two weeks and the team shipped eleven times. That had never happened.”
- Engagement
- Survey → Platform → Operate, 9 weeks
- Stack
- TypeScript, GitHub Actions, Terraform, Kubernetes
- Team
- Two from Intervia, three from Sable
- Handover
- Runbooks, two training sessions, 30 days of cover
Gallery