This study experimentally investigated the properties of sandstone concrete, focusing on the effects of silica fume content at ratios of 10, 20, and 30 %, along with six compaction levels: no compaction, light compaction, heavy compaction, and vibrating compaction for 10, 20, 30, and 60 seconds. This study introduces a machine-learning model to predict the compressive, splitting tensile, and rupture modulus. Eight machine-learning algorithms — Linear Regression, Decision Tree, Support Vector Machine, Efficient Linear, Ensemble Decision Trees, Gaussian Process Regression, Artificial Neural Network, and Kernel Models — were applied to the dataset, with hyperparameters tuned to create 28 total models. The performance was assessed using evaluation metrics such as RMSE, R2 , and MAE. The experimental results highlighted the importance of compaction methods and silica fume content on desirable concrete properties, with a maximum slump loss of 2.54 % observed in the mix containing 30 % silica fume compared to the control mix. In addition, the low density of silica fume mitigated overcompaction, achieving adequate density after 20 seconds of vibration across all the mixes. Samples with 20 % silica fume and 10 seconds of vibration showed significant improvements in the compressive strength (25.7 %), splitting tensile strength (52.4 %), and modulus of rupture (153.8 %) compared to the control sample. The machine-learning findings suggest that specific models can reliably predict the mechanical properties. Certain configurations of Linear Regression and Support Vector Machine techniques effectively predicted the mechanical properties of controlled-compaction sandstone concrete with R-squared value up to 1.0, showcasing the potential of machine learning to optimize concrete mix designs despite limited real-world data