Developed binary data encoding support for the mosaicml/streaming repository, focusing on enhancing the dataframe_to_mds converter to map Spark BinaryType columns to binary-encoded MDS types such as PNG and JPEG. Leveraged Python and data engineering skills to implement robust schema mapping and added validation logic, ensuring that only binary columns are eligible for encoding with these formats. This approach improved the flexibility and reliability of binary asset ingestion, particularly for image data, within MDS-based storage and analytics pipelines. The work addressed the need for flexible data representations and strengthened data validation in complex data conversion workflows without introducing new bugs.