• Designed and implemented an event driven Pub-Sub framework intended for company-wide adoption, which aim to ease the breakup of the old monolithic codebase into microservices. Redesigned and launched the new payment system built on top of this framework, boosting the traceability and data consistency of the system

• Developed various reconciliation workflow and compensation workflow to boost the accuracy of the ledger system, driving down the discrepancy between actual disbursement amount and calculated merchant balance from 1 million to at most 1 hundred

• Led the ledger error reduction initiative, identified and fixed various issues, reducing the overall error rate among all payables from 3% to lower than 0.1%

• Delivered features and improved system performance, including reducing account balance api latency by 70%, automating payments objects creation during the disbursement process, etc.

• Designed and launched Nike shoes notification system in Japan region, which notifies the subscribed users when their interested shoes became in stock. After launch this application attracted ~5k DAU, almost 3x expansion of the user base

• Designed and launched the POI (place of interest) system in Singapore region, which includes poi tagging workflow and poi landing page, enabling the users to search places and operation/product team to promote campaigns/events around places, increasing the user retention rate by 4% on weekly basis

• Setup the workflow for aggregating POIs that are frequently liked or bookmarked over certain time windows, helping operation/product team decide which ones need attention

• Implemented a timer-based card borrowing incentive, allowing players to borrow time-constraint cards into their card decks

• Implemented a user inbox system, where if a user watched card become eligible to borrow, users can see a message in their inbox

• Implemented an email notification system for event types like when a user is challenged by others to play a game, a watched card is eligible to borrow, extra tokens is distributed etc.

• Developed and launched ControlPlane (the services manage all clusters lifecycle) features such as allowing running DAG on clusters, auto scaling nodes based on certain metrics, auto replacement of unhealthy nodes, allowing running tasks in parallel etc.

• Developed InstanceController (the daemon service run on each cluster) features, identified and fixed various multi thread issues like deadlock, threads blocking each other, single thread stuck etc. which leads to service unavailable

• Implemented a one-click workflow to generate regionalized credential and automation scripts to accelerate building EMR services in new AWS regions, reducing build time from 3 weeks to 1 week