During October 2025, Daicong worked on the zipline-ai/chronon repository, delivering a feature that enables configurable Kafka offset starting points for Flink streaming jobs. By allowing users to specify a custom start timestamp, Daicong’s solution supports historical data backfill, reprocessing, and efficient recovery after downtime. The implementation in Scala included robust timestamp validation and error handling, ensuring that invalid formats are caught early. For backward compatibility, the system defaults to the latest committed offsets when no timestamp is provided. This work demonstrates depth in Apache Flink, Apache Kafka, and backend data engineering, addressing real-world challenges in streaming data pipelines.