Developed and delivered a new feature for the scikit-learn repository, introducing monotonicity-constrained tree models with enhanced missing value handling for dense datasets. The work enabled decision tree classifiers and regressors to enforce monotonic constraints, addressing requirements for safety and trust in regulated domains. Using Python and leveraging the scikit-learn codebase, the implementation improved model reliability and reduced the need for extensive data preprocessing. The contribution involved close collaboration through pull request workflows, code review, and testing practices. This addition strengthened the robustness of tree-based models, making them more suitable for applications where data integrity and interpretability are critical.