Designed and implemented a attribution gap mitigation for Chrome 3P cookie phase-out (Google PAA) in Amazon DSP measurement, reconstructing ~3M daily conversions and supporting reporting continuity.

Built distributed Spark pipelines on EMR with exact-once guarantees, implementing traffic splitting, deduplication, and id-less attribution logic to ensure correctness under evolving privacy constraints.

Drove data lake migration from legacy S3 Hive datasets to Apache Iceberg, reducing ~$2.3M/year in storage and compute costs while significantly improving query performance and data reliability for DS and ML teams.

Designed shared Spark–Iceberg read/write frameworks with schema validation, partition controls, and safe overwrite semantics, enabling scalable restatements, backfills, and long-term table maintenance.

Improved system scalability and operational excellence through dashboards, alarms, EMR auto-scaling automation, and on-call SOPs; collaborated closely with PM, Data Science, and ML engineers to deliver under tight timelines.

Introduced a metrics-driven concurrency controller to manage hotspot shards and uneven event distribution.

Contributed as a second inventor on two AWS patents related to dynamic shard-level parallelization and concurrency for event streams.

Researched and organized teaching materials for five faculty members, supporting effective learning.

Coordinated with senior professors to facilitate postgraduate research dissertations, enhancing academic collaboration.

Developed data manage tools to synchronize timestamp field form different hardware, such as H10, OH1

Configured and maintained Jenkins which provides automated and continuous delivery of the KPI report

Developed efficient script-driven backfill jobs using SQL and Python, optimizing batch sizes for seamless migration.

Implemented rigorous data validation checks to ensure accuracy before and after the cutover.

Coordinated the final cutover with minimal service disruption, ensuring a smooth transition for production operations.