Worked on the apache/hudi repository to enhance the reliability of data ingestion pipelines by addressing a bug in the DebeziumSource component. Implemented a fix in Java and Spark to ensure that a DataFrame is always emitted with a schema, even when no new messages are present, thereby preventing empty results and downstream failures. Developed targeted unit tests to validate schema presence in no-new-messages scenarios, improving test coverage and reducing the risk of regression. This work stabilized the integration of Debezium and Kafka within Apache Hudi, ensuring consistent schema handling and minimizing error conditions in downstream data engineering workflows.