I propound that, in essence, there are two ways in which our pipelines can break: through breaking code changes and through “bad data” extracted from a raw source. In this talk, I will argue that these can both be remedied with a GitOps approach to data.

We will dive into what GitOps for data should look like and investigate how it would overcome the two root problems for pipelines. We will then investigate a solution that bridges the existing divide between data and code: Virtual Data Builds (VDBs).