Ads Callback Platform

➤ Developed and maintained distributed stream processing pipelines that ingest ads impressions and engagements in real-time for downstream use cases like Ads Serving, Billing, and Prediction Model.

User Data Platform / Membership Platform

➤ Designed and implemented large-scale stateless RESTful Profiling API, which is a centralized Profiling Data Ecosystem, provides data to customers across the company, and serves 150K+ QPS globally

➤ Designed and built a secure data ingestion pipeline to consume customer’s Profiling data from Yahoo Grid; the ingested Profiling data is served as the source of multiple company-wise backend services. The new design pipeline reduced onboard of new Grid feeds from weeks to days

➤ Maintained and implemented scalable, replicated, and low-latency key-value NoSQL user database service which contains 3B+ user accounts and serves 3M+ QPS globally

➤ Was the leader of team’s service operation maintenance, including keeping packages and platforms update-to-date, building configuration management tools and CI/CD pipelines for service deployment on 1000+ hosts, and providing good documentation and an easy-to-understand operational model

Identity Platform

➤ Implemented a stats-logging-service API which handles 1B+ HTTP requests per day for Yahoo login service and saved 30% man-hours in stats logs analysis

➤ Wrote 100% coverage unit tests and integration tests for the API using JUnit/JMockit and Cucumber

➤ Utilized Splunk to analyze accounts activity and performed data exploration to identify login pattern and prevent from hacking activities using distributed computing (Hadoop, Hive, and PySpark)

Project: Password Security Guard

➤ Implemented a new web service serving 700M+ global accounts including Yahoo, AOL, etc, to prevent 50% users from using breached passwords

➤ Collaborated with UI designers and implemented grid layout for the new change-password page which adds easy-understanding instructions, saving user 20% time on the page, based on Nodejs architecture and developing/testing on Docker

➤ Built a robust service server using Undertow and RESTful API with SSL certificates

➤ Integrated codes between Yahoo login service stack, frontend stack, and this new service, including unit tests and integration tests

Design Signoff Team

➤ Developed various machine learning algorithms to build predictive analytics metrics to improve chip performances. The scalable solution also reduced run time by 20%

➤ Automated 5+ workstreams in the verification phase for IC design using Python-based high-performant algorithms. My work resulted in a 30% performance increase and was standardized into the workflow at MediaTek adopted by more than 50 engineers

➤ Managed 3 crossed-team smartphone projects, and applied a more data-driven approach into IC design verification, static timing analysis, dynamic power analysis, yield analysis, and failed chip troubleshooting

Project: Dynamic Power Prediction

➤ Developed a machine-learning solution for IC chips to predict power fluctuation under 10% error rate with a 7x time performance increase

➤ Trained multiple neural nets and Xgboost Python-based models