Worked on Apache Druid to deliver a critical fix for Delta ingestion from Parquet sources, addressing a bug that previously caused incomplete row ingestion from multi-file Parquet inputs. The solution involved promoting the per-file iterator to a field, ensuring all batches were drained before advancing, and properly closing iterators to prevent resource leaks. Regression tests were added to validate complete batch processing, including scenarios with large row groups and multiple files. This work, implemented in Java with a focus on back end development and unit testing, improved data accuracy, reliability, and maintainability of Delta-based Parquet ingestion in the repository.