Bio

For the past few years, I have worked at the intersection of Compute platforms, Continuous Deployment (CD), and later Engineering Effectiveness (Developer Experience). As part of the Compute team, our philosophy was that aside from the agility benefits of CD, we can increase the adoption of managed Compute offerings if we lower the barrier to setting up CD for our customers. We designed the CD tooling (a generic pipeline execution engine) as the common interface to all the Compute platforms. Specifically, we built the CD tooling from the ground up to hide many of the peculiarities of deploying to Aurora/Mesos and Kubernetes. We applied a similar approach to help DevEx teams change the build tooling from Pants to Bazel. I built a lot of integration points in our deploy tooling with the rest of the Twitter infrastructure stack, like Kubernetes, observability and monitoring, notification systems, code review, and a few other safety tools we ran internally. This helped teams move to an automated commit to production deployment without any human interactions. While the pipeline executor was greatly tuned for production deploys to Compute platforms such as K8s and Aurora/Mesos, it was generic enough to handle other use cases such as safe deployment to bare-metal, machine learning pipelines, and performance tuning and testing.