Developed the Checkpoint Process Metadata Persistence feature for the google/orbax repository, enhancing the checkpointing system to save and restore distributed process metadata alongside checkpoint data. This work involved refactoring the existing system to introduce a dedicated process metadata handler, ensuring that distributed process information is preserved and enabling accurate reconstruction of mesh configurations during restoration. By addressing the challenge of maintaining distributed state, the solution improved checkpoint robustness and reduced manual recovery effort. The implementation leveraged Python and system design principles, with a focus on distributed systems and checkpointing, resulting in more reliable and maintainable distributed training workflows.