- Led the Controller Manager team in architecting and developing a new framework for Kubernetes controller to enhance operability, incorporating advanced instrumentation for logging, tracing, debugging, and metrics monitoring. It delegated routine logic responsibilities, such as finalizer management and resource specification, to streamline developer focus on essential business logic. Furthermore, it orchestrated the transformation of a monolithic reconciler into a constellation of modular feature modules, vastly enhancing testing efficiency and promoting seamless cross-controller module integration. - Enforced and upheld conventions for logging, metrics, and conditions to foster standardized practices. Led the design of linters instrumental in elevating the overall health of the codebase across the entire organization and effectively maintaining uniformity and adherence to standards. - Successfully influenced and contributed to over 200 controllers within more than 20 controller managers.

Trust Experience - Built and maintained several different friction flows for China users to help them complete account ownership or captcha verifications, improving the user experience for the whole flow as well as blocking library hit and account takeover attacks from black industries - Built the whole flow for China business accounts registration onboarding process, including storing registration data to database, assisting to build registration QA review dashboard and sending emails or in-app action cards once approval or denial. Listing Verification - Built and maintained China listing verification service to detect amenity, room type, and any potential redline information from the text and images uploaded by users, kept tuning model to increase the detection accuracy. It largely reduced the cost from manual review and improved the overall efficiency of the listing verification process.

Growth Team (Built entire offline postcard channel to resurrect dormant and inactive users) - Develop user content generation, process backend logic with personalized discounts, then schedule the campaign to run routinely as a cron job - Run campaign experiments, aggregate experiment results, maintain data pipelines, and then build visualizations (Looker) that analyze project revenue or user engagement - Leverage insights from those visualizations to recommend growth strategies, implement cost-saving practices, or localize content Help to develop online email and push notification channel - Maintain stability of email and push notifications across different platforms, while optimizing the frequency of millions of daily sends - Be responsible for the scalability of communications by testing HTML emails with Litmus technology and push notifications on iOS and Android devices

The snack contributor of Story backend team The bug maker and fixer in the team Projects: Async logic for media variant transcoding and replication to S3 and GCS

Funded by Palantir and Credit Suisse, Signac leverages the power of Big Data, Data Science, and Artificial Intelligence to help financial institutions better manage risk, and prevent rogue traders and other employees from engaging in activities that endanger our financial system.

- Developed data cleaning code to preprocess data and built formatted datasets for predictive analysis of assembly line downtimes - Designed cost function based on sensitivity and specificity measurement and tuned parameters according to the business goal - Implemented several time series models on large scale datasets to validate whether time series analysis worked well on the problem - Implemented unsupervised learning methods to train different models and compared the prediction results by ROC curve

- Built a dashboard for the Medical Science Liaison (MSL) team understand where existing relationships with Key Opinion Leaders (KOLs) are in place, and the type of activities involved in those collaborations - Implemented supervised learning methods to build the model for predicating the optimal treatment of individual epilepsy patient - Receiver Operating Characteristic (ROC) was applied to examine the best prediction model and prediction rate reached up to 85%