Developed a Maximum Mean Discrepancy (MMD)-based Distribution Discrepancy Operator for the OpenDCAI/DataFlow repository, enabling quantification of distribution differences between embeddings of evaluation and reference datasets to support data-drift detection and enhance evaluation reliability. The work involved Python operator development, leveraging data analysis and machine learning techniques to implement auto-checks with lazy loader imports for streamlined validation and optimized startup. Initialization was refactored by relocating key parameters to the operator’s constructor, improving reproducibility and startup speed. The operator’s documentation was updated for clarity, laying the foundation for deeper analytics and continuous integration within model evaluation pipelines. No bugs were reported.