Passing tests and caching checks.
2021 — Now
2019 — 2021
Supported a new product without increasing operating costs by spearheading a seamless modification to our Python GraphQL execution engine that allows us to discover requested fields and prune requests by 3x before querying our microservices.
Simplified the process of discovering marketable wins by constructing a Reactive Angular UI component that gives users more confidence when choosing and visualizing their data.
Attained a 200% increase in user engagement by leading a team of junior and senior engineers to implement a new set of UX-friendly data templates in Angular/Typescript.
2018 — 2019
Mountain View
Optimized our Python data pipelines to prune incoming data sources more intelligently, while reducing operating costs, to deliver crucial metrics 6x more frequently to our customers, allowing them to quickly market wins about their digital content.
Increased efficiency of our Data Science team by collaborating cross functionally on a technical plan that enables them to continuously integrate their new metrics into our pipeline and allow seamless experimentation and visualization.
Developed a modular framework in Python that allowed my team to quickly create metrics for monitoring and validation, capable of halting the entire pipeline and alerting team members if failures occurred.
Mountain View
Increased the number of known Youtube user ages by 33% by using Spark and Regex to mine through all Youtube comments and user biographies to discover over 2.1 million user demographic data points.
Used Python to create a CLI that allows developers to view and/or cache the top requests for any API endpoint.
Utilized Elasticsearch to examine the API request logs and retrieve the top requests based on some user specifications, such as the top most frequent, failed, or high latency requests.
Ensured the consistent use of accurate data in production by creating a verification pipeline for pushing newfound data, such as those found during data mining, from development to production databases.
Singapore
Reduced map-matching computation time by utilizing the Even-Odd Rule to develop a program that determines whether a GPS coordinate is within a geographical polygon. Thus, the map matching algorithm only computes on the data points within that polygon rather than the whole data set.
Utilized Java to create a Hidden Markov Model based map-matching algorithm to achieve an increase from 85% to 94% in average accuracy matching GPS coordinates to road networks.
Launched a server API using Java Spring to expose the map-matching algorithm and real-time traffic updates to mobile devices.
Employed JSON Web Tokens to secure and authenticate API requests
Visualized geo-spatial data and map-matching results using QGIS.
Education
2014 — 2018
University of California, Berkeley
Bachelor of Arts (B.A.)
2014 — 2018