This study delves into understanding the relationship between synthesis conditions and the resulting properties of the Zr-based metal−organic framework (MOF) UiO-66, with an emphasis on machine learning (ML) in making quantitative predictions. Utilizing a comprehensive, manually curated data set, three ML models are trained to predict UiO-66 properties, including specific surface area, defect concentration, and particle size, based on synthesis parameters. A solution to the inverse problem, which involves finding optimal synthesis conditions for given properties using the method of differential evolution, has been implemented in the software. Experimental validation of models through synthesis and detailed characterization of UiO-66 samples and comparison with the predicted properties show a high accuracy, confirming their reliability. Interpretation of the ML models using Shapley additive explanation values and twodimensional (2D) partial dependence plots reveals both previously known patterns, validating the adequacy of the models, and new, previously unexplored patterns in the relationships between the synthesis conditions and UiO-66 properties. The developed models can be used as a basis for further research on MOF synthesis. This approach can be applied to the rational design of UiO-66 for various applications.