Worked on the keras-team/keras-hub repository to implement PaliGemma2 mix checkpoints and presets, supporting 3B, 10B, and 28B model configurations with image resolutions of 224 and 448. Leveraged Python for scripting and configuration management to update convert_pali_gemma2_checkpoints.py, enabling seamless mapping and integration of these new checkpoints into existing workflows. This addition improved deployment flexibility and reproducibility by making configurable checkpoints available in release artifacts and documentation. The work focused on enhancing model integration and readiness for downstream fine-tuning pipelines, aligning closely with project roadmap objectives and demonstrating effective collaboration within the machine learning development process.