Working on Standardization and Automation of our Kubernetes infrastructure. Previously embedded with the Stream Processing Platform team utilizing Apache Flink.

Production Engineering at Facebook is a hybrid between software and systems engineering; it keeps Facebook running smoothly and scaling efficiently. Worked from the Menlo Park, CA headquarters from 2016-2021. Moved to permanently remote in Austin, TX since July 2021.Worked in Data Infrastructure/Warehouse with Hadoop, HDFS, Hive, Spark, Presto etc. on a few teams (most recent first):Global Tetris Router:- Load balancing compute queries into regions, datacenters, and storage clusters while considering: - Data locality (where data is read from, where data will be written to) - IO limits within our storage clusters - Query backlogs + cluster sizes for compute engines- Changed the API from routing to a particular datacenter to aggregating metrics and routing per region. This allowed us to realize efficiency gains by deploying our storage clusters regionally, reducing the required spare servers at the cost of more networking bandwidth usage.- Updated SLI/SLOs as we found production issues not covered under our existing SLI/SLOs. Ensured that customer pain is captured within our top-level metrics and alert us before user reports.Warehouse Shared Infra:- Took ownership of all of the hosts/servers running Data Warehouse workloads to centralize low-level settings and management (kernel upgrades, misbehaving/faulty servers) using tools including Chef. This allowed us to pool together buffer capacity, reducing total server demand.- Deprecated the Data Warehouse host pool as Facebook centralized host ownership company-wide, ensured our requirements were met in the new pool and helped teams migrate.Dataswarm:- Query scheduler that waits for data to become available before executing (similar to Airflow)- Enabled deprecation of our old query scheduler by implementing an Operator that dynamically schedules Operators at runtime rather than at build time.

Worked in a team of 5 to speed up application launch time on Android devices with Qualcomm’s SoC:-Developed a network tracing tool (in Python) to evaluate networking’s impact on application launches-Integrated my Python network tracing tool into an existing Python analysis tool as a plugin-Researched and evaluated different caching methods in C to improve data access times-Researched the Linux kernel’s function calls involved with TCP connections and added relevant traces

Added features and fixed bugs to Web Forms ASP.NET multirole websites with Microsoft SQL Server:- Implemented dependency injection with Unity- Modified SQL database to enable new features and create new relationships between data- Encrypted sensitive user information in the database- Extended functionality on judging systems and database reports for adminsWorked part-time throughout the school year.