One Pipeline
CI/CD platform serving 20,000+ engineers at Capital One
The Problem
At Capital One, thousands of engineering teams were running their own CI/CD configurations — each slightly different, each requiring its own maintenance, and none talking to each other. Build times varied wildly. Security scanning was inconsistent. The cognitive load on developers was enormous.
The question wasn’t whether to standardize — it was how to do it without revolting 20,000 engineers.
The Approach
We built One Pipeline: a centralized, opinionated CI/CD platform that makes the right thing the easy thing. The core insight was that most teams don’t want to configure CI/CD — they want CI/CD to work.
Key design principles:
- Convention over configuration — sensible defaults that cover 90% of use cases, with escape hatches for the remaining 10%
- Zero-diff migration — teams could onboard without changing their application code
- Visibility first — a unified dashboard showing build health, coverage trends, and deployment status across the portfolio
Scale
One Pipeline currently processes over 8 million builds annually across Capital One’s engineering organization.
The Outcome
- 92% of GitHub repositories now on One Pipeline
- 8M+ annual builds processed through unified infrastructure
- Significant reduction in build configuration maintenance burden per team
- Consistent security scanning across all repositories — a compliance win
What I Learned
Building internal platforms is fundamentally a trust problem. Engineers don’t adopt tools because they’re told to — they adopt them because the tool is clearly better than the alternative. The first six months were spent building credibility: fixing rough edges fast, listening obsessively to feedback, and shipping improvements weekly.
The other lesson: platform teams are product teams. We ran discovery interviews, maintained a public roadmap, and held quarterly office hours. Treating our internal users as customers made all the difference.