• Improved scalability of GDPR compliance systems that evaluated over 500 billion records weekly. - Removed artificial concurrency limits in Airflow DAG structure, replacing them with flexible parallelism limits using Airflow task pools. - Resolved EMR cluster scale problems by migrating jobs to the Instance Fleets API and using cheaper and easier-to-schedule instance types.• Helped reduce team’s average weekly alerts by 50% by implementing quarterly operational excellence reviews and proposing and implementing improvements. - Collaborated with other team members to implement as many of the proposed improvements as possible in the time allotted.

• Built and maintained analytics pipelines for search engine user engagement and content quality.- Built and optimized Spark jobs to join millions of search, user engagement, and advertising logs to build combined datasets used to optimize search results and increase user engagement.- Optimized Looker data models to reduce frequently used dashboard refresh times from over 5 minutes to under 1 minute.- Extended a text-based search results comparison tool to automate recurring comparisons, store the results in a SQL database, and visualize the results using Looker.- Collaborated with the search content team to plan and implement a quarterly roadmap of content quality metrics and insights.• Migrated ad attribution logic and data, including over 50 billion historical records, into the Branch attribution systems to support a product acquisition and customer migration.- Built Spark jobs to filter out unnecessary data from billions of attribution logs, reducing the overall import data volume by over 30%.- Proposed and implemented an on-call review strategy that exposed common failures of the high-availability ad attribution pipeline, resulting in a sustained reduction in on- call pages.

• Subject-matter expert and technical leader for the Go programming language at TUNE.• Owned the migration of a high-volume web service from AWS EC2 hardware to shared and autoscaled AWS ECS hardware, saving over $20k/month.- Developed a local disk buffering library that allowed running multiple service instances on the same ECS host.- Diagnosed and resolved a serious performance issue by replacing a SQLite dependency with a native-Go in-memory database.

• Owned development and deployment of an AWS SQS message packing library that reduced SQS API call frequency by over 50%, saving over $20k/month.• Coordinated the application update component of a data deletion project that reduced DynamoDB storage cost by over $200k/month.

• Took on team management responsibilities while the team had no dedicated SDM and established a Scrum-like iterative development process

• Researched Go scheduler performance, focusing on avoiding the pitfalls of Go's semi-preemptive scheduler, and delivered findings in a presentation at a Seattle Go Meetup• Established TUNE’s initial Go development practices and helped build TUNE’s first production Go service

Created and supported a video content ingestion and tracking platform for Amazon Instant Video, capable of ingesting and analyzing over 150TB of video content per day.• Automated deployment to a high-availability fleet running Aspera Enterprise Server, including graceful management of long-running video transfer processes• Updated the encoding service to fetch a frequently-changed configuration file from AWS S3, saving 6 dev hours and over 300 encoding fleet hours when adding content partners

Developed features for and supported Vitrue's Publisher product, a Ruby on Rails social media marketing automation and management cloud application.• Spearheaded creation of a web API for Publisher to support a new mobile client• Selected for a team to integrate Google+ resource management into the Publisherproduct- Collaborated with the Google+ Brand Pages API team to troubleshoot and- improve the Brand Pages API before it was publicly available