During July 2025, this developer delivered three targeted enhancements across the metatensor/metatrain, lab-cosmo/atomistic-cookbook, and metatensor/metatensor repositories. They introduced field-level data loading in Python for DiskDataset, optimizing memory and I/O by allowing selective field retrieval. In lab-cosmo/atomistic-cookbook, they improved CI/CD workflows using GitHub Actions and YAML, enabling authenticated artifact access for forked pull requests and resolving API rate limit issues. Additionally, they enhanced tensor operations in metatensor/metatensor by adding an option to skip gradient metadata checks, streamlining large model comparisons. Their work focused on API design, dataset management, and robust unit testing to improve efficiency and reliability.