Worked on the apache/pulsar repository to refactor the Kafka Connect source offset flushing logic, focusing on improving reliability and efficiency for high-throughput data ingestion. Leveraged Java and asynchronous programming techniques to introduce a centralized flush trigger and a new mechanism for managing flush futures, reducing complexity and minimizing race conditions. Expanded test coverage to include scenarios such as message filtering and partial acknowledgments, ensuring correctness under varied operational conditions. The changes aligned with Pulsar IO’s stability objectives, enhancing offset management and reducing operational risk for Kafka Connect users while strengthening data integrity across distributed ingestion pipelines. No bugs were reported or fixed.