Is Your CI/CD Pipeline Actually Working?

Satya Singeetham (EVP and CTO @ New Horizon Bank) spent 17 years at FIS building CI/CD from scratch for a core banking platform running across up to 15 engineering teams. A bad release in that environment does not just break a feature. It stops a bank from opening its doors the next morning. On this episode of 1 IDEA, Suresh Mathew sits down with Satya to unpack why CI/CD must be built as an engineering product and what it takes to make release a repeatable, non-negotiable process at scale. We cover: Why CI/CD must be a product, not a tool or social contract The release train model: fixed cadence, no exceptions Quality gates, SonarQube, and Docker builds at scale What a bad release costs in a regulated banking environment T minus 2: backward compatibility across 10 to 15 versions Where AI fits inside a disciplined release control system CHAPTERS: 00:00 Introduction 00:31 The FIS monolith: 300 banks, one giant COBOL core 03:05 Rebuilding from scratch: componentized banking platform 04:35 The CI/CD challenge with 8 to 15 parallel teams 06:18 Treating CI/CD as a product, not a tool 08:34 Quality gates, Docker images, and automated deployments 13:01 The real cost of a bad release in banking 19:11 Release sequencing, dependency problems, and Flyway 22:31 Building the culture: education over enforcement 29:34 How AI changes the CI/CD equation