Fail Fast & Ship It with Jeremy Custenborder | Ep. 18

Jeremy Custenborder’s journey began in the late 1990s, when he was a self‑taught network administrator for a county library and later a developer on MySpace. With no off‑the‑shelf caching or orchestration tools, his team built proprietary solutions on Windows and ASP.NET 2.0, handling tens of millions of users and terabit‑scale traffic. Those constraints forced a ‘fail fast, iterate quickly’ mindset, turning every outage into a learning opportunity. The experience of manually imaging servers, stitching together NetScalers, and pioneering early load‑balancing techniques laid a foundation for modern distributed‑system thinking.

At Confluent, Jeremy translated that hands‑on expertise into the early Kafka Connect ecosystem. He authored the first JDBC and HDFS connectors, establishing a pattern for pluggable data pipelines that abstracted source systems behind a unified schema. By championing Avro serialization and the Schema Registry, he gave developers confidence in data contracts across microservices. The open‑source connector marketplace he helped seed provided vetted, reusable components, accelerating customer adoption of Apache Kafka for real‑time streaming. His pragmatic approach—shipping working code, then refining it—mirrored the same fail‑fast philosophy that had served him at MySpace.

Jeremy’s story underscores three timeless lessons for today’s data engineers. First, building custom tools is acceptable when the ecosystem lacks mature options; the resulting deep SRE knowledge pays dividends later. Second, embracing rapid iteration and continuous feedback reduces technical debt and speeds time‑to‑value in high‑throughput environments. Finally, contributing to open‑source projects like Kafka Connect creates communal value and establishes industry standards. As Kubernetes, Prometheus, and managed cloud services dominate, the underlying principle remains: start simple, fail fast, and let the community refine the solution. Engineers who internalize this mindset can navigate the ever‑evolving data‑streaming landscape with confidence.