Zizhun Guo

- Built ZhunWorld and its self-hosted content data platform, turning user interaction data into real-time, hourly, and daily homepage trending rankings. - Built an event pipeline with frontend tracking, Django, Kafka, and S3 to collect interaction data and ingest partitioned raw data into the lake. - Developed Airflow workflows and dbt models on Redshift to orchestrate processing, clean and deduplicate events, and generate engagement-based ranking outputs. - Implemented GitHub Actions CI/CD and a Redis-back

- Designed and built a Python/SQL batch processing application on a Hadoop (Spark + Hive) platform to deliver daily analytics datasets for three online game products with a combined monthly gross revenue of ~RMB 50M. - Improved daily batch runtimes by roughly 75% through parallel execution of Spark/Hive jobs and Spark SQL/query optimization. - Authored and maintained SQL across 100+ analytics tables to produce cohort and retention datasets and reports for operations teams.

- Trained and deployed supervised and reinforcement learning models as services for player–AI matchmaking and Mahjong AI agents, improving matchmaking quality and in-game AI strength. - Implemented PySpark/Hive batch jobs to build T+1 user-behavior feature tables in the data warehouse, serving as feature stores for matchmaking and game-AI models. - Contributed Java-based feature parsing logic to a Kafka/Flink/Redis streaming component, working with data engineering and platform teams to integrate feature pi