Developing something new - Python, React-Native, Go

Part of Messaging Infrastructure:

• Worked on both Messenger features and Instagram Direct Messaging features.

• For Messenger: Admin features for Community Chats and View-Only Chats. Read about Community Chats: https://about.fb.com/news/2022/09/community-chats-on-messenger-and-facebook

• Instagram Direct: Migration of IG direct endpoints to Messenger Architecture PHP, including API design. Validated via shadow testing, A/B testing, and gradual rollout.

• Led reliability initiatives like testing best practices designs.

Also worked with team to design and implement Subscriber Chats for creators to message followers: https://about.fb.com/news/2022/07/new-ways-for-creators-to-connect-with-subscribers-on-instagram

• Optimized BigTable usage by integrating with the Spotify BigTable Autoscaler

• Worked on a next generation Analytics pipeline. This was done through the use of Beam Pipelines which would take inputs from BigQuery or PubSub and write it to either Postgres, BigQuery, or another PubSub topic. This ran in Dataflow.

• Helped manage and maintain Airflow for Data Scientist use cases.

• Optimized query usages for a cost- intensive portion of our application.

• Wrote a conversion from business-specific Python code to Go code to be more performant.

I worked primarily in the Google Cloud ecosystem to develop highly scalable and distributed software. A big portion of my time was spent migrating Python2.7 code from the Google Application Engine Monolith to either Python3.7 or Golang code in the Google Kubernetes Engine as a microservice. In doing that I learned how to create services that are easy to monitor which are capable of handling hundreds of thousands of requests per second. This was prioritized because of a need to reduce cost while also being able to increase throughput for the same workloads. I also learned how to safely and efficiently migrate from the monolith, or deprecated system in general, to newer systems.

I also led a project that standardized error handling and retries across the organization. The goal of this was to make error handling safe and more resilient to changes in an API agreement from a service. I wrote a blog post, which I linked, describing the problem and the solution.

More recently I have been working on larger scale cost initiatives, one of which helped reduce infrastructure cost about one million dollars a year. My role in this project was in creating a validation process and tooling that would speed up the validation. These tools made validation very easy while not lowering quality and drastically sped up the migration itself.