Worked on the zarr-python repository to enhance metadata robustness for Zarr Format 3 by addressing the encoding of special floating-point values during consolidated metadata serialization. Developed a Python-based helper function that replaces NaN, Infinity, and -Infinity with their string representations, ensuring compatibility with ArrayMetadata behavior. This approach improved the reliability and cross-compatibility of metadata handling across the Zarr ecosystem. The work focused on bug fixing, data serialization, and metadata handling, with thorough testing to validate the solution. The update resolved a key issue in consolidated metadata encoding, contributing to more robust and accurate data management within the project.