The management of hypersaline wastewaters, including desalination brines, poses significant environmental and technical challenges. Temperature swing solvent extraction (TSSE) offers a promising solution by exploiting the temperature-dependent water affinity of low-polarity solvents, typically amines. In this study, machine learning (ML) is applied to enhance TSSE by predicting the mutual solubilities of water and amines. Experimental data and amine properties were used to train eight ML models. A stacking ensemble strategy yielded high predictive accuracy, with R2 values of up to 97.8% for water-in-amine solubility and 95.8% for amine-in-water solubility. By integrating ML predictions with multi-objective optimization, the study identified ideal amines and operational temperatures.