The rising urgency to build climate-resilient infrastructure, particularly in the education sector, has underscored the need for accurate cost prediction tools that support sustainable decision-making. This study proposes a Gene Expression Programming (GEP)-based predictive model to estimate the cost of constructing green resilient school buildings in hot desert climates. The data were collected through structured, on-site questionnaires and interviews with 3687 architects, engineers, and other professionals involved in school infrastructure projects in Bahawalpur and Multan, emphasizing aspects such as design approaches, material selection, energy efficiency, and climate-responsive construction methods. The compiled dataset underwent comprehensive preprocessing, which included data cleaning, normalization, categorical variable encoding, and outlier removal. Through expert consultation and statistical correlation techniques, 19 key predictors were identified. Results show that the GEP model delivers high predictive performance R-squared (R2 = 0.81), Root Mean Square Error (RMSE = 0.012), Mean Absolute Error (MAE = 0.023) and Mean Square Error (MSE = 0.013) outperforming Decision Tree, SVM and Hybrid models and demonstrating robust generalizability while explaining 81 % of the dataset variance. Feature importance shows that lack of awareness and cost of operations contribute the most to model development. The GEP model was further validated with independent data of 793 professionals (R2 = 0.76), and a Taylor diagram was used to compare with the existing models while generating an interpretable mathematical architecture and expression for cost estimation. The study highlights the potential of evolutionary algorithms like GEP in stakeholder understanding in sustainable school infrastructure projects, thereby contributing to climate adaptation.