Architected and built an enterprise-level data center from 0 to 1, including Hadoop-based data infrastructure, data integration, ETL pipelines, data workflow, data warehouse, data governance, and BI, ensuring system availability ≥ 99.99%

Standardized data query layer with unified SQL, compatible with Hive, Spark, Presto/Trino, Elasticsearch, and HBase. Boosted data development and usage efficiency by over 50%

Built and optimized data warehouses by applying data modeling best practices and a multilevel data model, achieving data harmonization, increasing data usage by over 10%, and reducing average query time by 15 seconds

Established data and integration standards for server logs and client SDK, simplifying multi-project data integration and reducing communication efforts by over 70%

Implemented multi-tiered expert strategies utilizing over 50 user dimensions to detect in-game cheating, resulting in savings of over $1 million for a single activity (25% of total costs)

Leveraged a combined DSSM model and expert strategy to provide in-game prop and activity recommendations, boosting the game's overall profits by over 10%

Performed data mining to identify potential causes of in-game recharge failures on Android and iOS and collaborated with game developers to determine the root cause, diminishing the recharge failure rate from 2% to 0.1%

Collaborated with the operations team and developed a multidimensional iOS refund model to monitor App Store refund issues, decreasing refund rate from a peak of 0.5% to 0.02% and saving over $20,000 in monthly profit loss

Employed HMM, NER, LDA, and BERT models, coupled with data visualization and real-time monitoring, to analyze multi-source user feedback, allowing project teams to understand users 10 times faster and with greater accuracy