Yashwanth Betha

Currently working in the Polaris platform team.

• Developed critical features for PACMan - our API server that we use to expose APIs over the internet - like setting up the ingress, creating a Python GRPC and a Go HTTP server and setting up authentication. I was also responsible for creating tools to create, delete and upgrade the infrastructure in which PACMan would be deployed. I also took complete ownership of writing E2E tests, setting up a Jenkins pipeline to run the said tests on a daily basis and fixing any issues that came up.
• Integrated PACMan with Swagger to auto-generate and expose documentation to end users, thus improving user experience.
• Improved existing PACMan infrastructure by moving the build process from a raw Dockerfile to bazel. This improved the build time from several minutes to few secondsm thus improving developer productivity.
• Developed a generate purpose Python metrics library that can be used to send arbitrary metrics to Signalfx. This library was used by PACMan to publish metrics like average latency per endpoint, number of failed/successful calls, etc.
• Developed a tracing library in Python that can be plugged into any GRPC server to track a request across multiple services.
• Led the design discussions for moving our existing poll based order provisioning system to a push based one. Currently mentoring an intern through the implementation phase of the project.
• Currently working on a capacity based licensing project, where we try to warn users when they try to use more capacity than they purchased. Led the design discussions and helped achieve consensus across multiple teams. Currently mentoring a fresh grad through the implementation phase of the project.

• Spearheaded the effort to move our Mongo based data aggregation pipelines to Apache Spark. This reduced our time runtime of our processes by almost 50%. I was responsible for all the phases of this project including the initial proposal, development, pushing it to production and validating it against our existing infrastructure.
• Improved our existing job scheduling infrastructure by breaking down large jobs into smaller jobs, introducing parallelism and by removing unnecessary dependencies between jobs.
• Worked on APIs related to generating distribution anomalies, timeseries anomalies and similar entities using numpy.