Jason Teoh Email & Phone Number

San Francisco, Ca, Us

Mountain View, Ca, Us

*Built multi-region, multi-cloud data ingestion processes for the data lake.*Built scalable Airflow pipelines across dozens of customers to aggregate aspect-level metadata into structured entity datasets, available for analytics consumption.*Worked with customer success team and customers to deliver standalone structured entity dataset process as well as data inquiries.*Planned long-term roadmap for data lakehouse including priorities, use cases, and risks.

San Francisco, Ca, Us

*Built new Beam and Flink infrastructure for on-premise streaming and batch computation pipelines within the Twitter tech ecosystem*Assisted the billing team in migrating use cases and developing scalable on-premises Beam-on-Flink solutions*Developed and supported configuration-driven streaming+batch aggregation and serving platform that supports critical revenue use cases (DataApi)*Managed analytics query server with pluggable backends (e.g., Bigtable, BigQuery) as well as general-purpose data loading workflows for loading data (Large Data Collider)

Los Angeles, Ca, Us

My research focuses on improving developer experiences with Big Data systems, with an emphasis on performance and correctness debugging. My prior works investigate how to enhance performance debugging for computation skew in Apache Spark and improve data provenance precision in dataflow applications through a combination of taint analysis and influence-based aggregation functions. More recently I have explored techniques which leverage targeted fuzz testing of individual user-defined functions in tandem with skew-inspired mutation operators to generate workloads which reproduce specified performance symptoms, particularly in scenarios where no suitable datasets are available.

Los Angeles, Ca, Us

CS143: Data Management Systems (Prof. Junghoo (John) Cho)Hybrid (in-person + remote) instruction for upper division data management course on how to store, manage, and access large-scale data using data management systems. Emphasis on RDBMS and SQL with an introduction to other systems such as Apache Spark and MongoDB.

Los Angeles, Ca, Us

CS 130: Software Engineering (Prof. Miryung Kim)Fully remote instruction for upper division software engineering course covering systematic engineering methods for large-scale software development: design and modeling methods; collaborative development environment; object-oriented design patterns and refactoring; build & integration; unit testing & regression testing; bug finding; software inspection and verification.

Los Angeles, Ca, Us

CS143: Database Systems (Prof. Carlo Zaniolo)Upper division database course with an emphasis on RDBMS and SQL and some introduction to more advanced systems such as Apache Spark.

Mountain View, Ca, Us

Designed and implemented an automated rule-based diagnostics framework to analyze query execution traces for performance problems and anomalies. Initial rules were designed based on collaboration with other Spanner developers, with room for enhancements and extension. Analysis results are integrated into an interactive query plan debugging tool for presentation and further debugging.

Sunnyvale, Ca, Us

(Continued LISTT/NCRM project from Software Engineering role until end of September 2015)ThirdEye: Open source anomaly detection framework for monitoring business metrics.* Utilizes Pinot (realtime distributed OLAP datastore with horizontal scaling and low latency).* Designed and implemented multifaceted dashboard for straightforward anomaly investigation.* Designed and implemented scalable anomaly detection framework capable of supporting anomaly history/lookback, feedback loops, distributed execution, email alerts, and configurable anomaly detection algorithms with a generic API.* Integrated dashboard and anomaly detection framework to provide a seamless user experience between anomaly detection configuration and investigation.* Onboarded use cases from data bootstrapping to server setup.

Sunnyvale, Ca, Us

LISTT/NCRM: Suite of self-service applications for computing custom member-levelattributes and targeting specific member audiences:* Enhanced computational flow to improve operational stability and data quality as LinkedIn grew to 400+ million members with over 500 metrics and dimensions per member in the data set coming from queries written in SQL (Teradata), Hive, and Pig as well as via user interaction on the web application.* Leveraged Datafu Hourglass to speed up key attribute computation and reduce delivery time.* Enhanced UI functionality to reduce business users’ turnaround time for audience segmentation such as segment upload and segment preview in real time.* Independently built flow monitoring CLI and graph UI to provide time and stability analytics on thecomputational flow as a whole.

Sunnyvale, Ca, Us

Backend development for self-service business analytics dashboard providing key metrics and dimensional splits. Designed REST API for frontend client interaction as well as an OLAP cube caching layer with background refresh daemon.

Assist residents with center resources and equipment, organize Game Nights (social events for residents, with food, board games, video games, and drinks).

Doctor Of Philosophy - Phd, Computer Science

Bachelor Of Arts (B.A.), Computer Science