Soil salinity is an essential constraint to sustainable crop production related to soil fertility, especially in arid and semi-arid regions. This study presents a data-driven approach for mapping soil salinity by integrating field-based electrical conductivity (EC) measurements with remote sensing and geospatial analysis in the district of Mandi Baha Uddin, Pakistan. Eleven georeferenced soil samples were collected and analyzed for EC (range: 0.59–1.06 dS/m), serving as training data for model calibration. Using Landsat 8 Surface Reflectance imagery within Google Earth Engine, spectral indices Normalized Difference Salinity Index (NDSI), Salinity Index (SI), and Brightness Index (BI) were extracted. Among various modeling approaches, a linear regression model was applied to these indices, revealing NDSI as the most significant predictor (coefficient = 3.862), while SI and BI showed negligible contribution. The model achieved moderate accuracy (R² = 0.526, RMSE = 0.089 dS/m). A Random Forest approach yielded higher training accuracy (R² = 0.811) but suffered from overfitting during cross-validation, indicating limited sample size constraints. The regression equation (EC = 3.862 × NDSI + 1.71) was applied in GEE to generate the EC prediction map. The resulting 30-meter resolution EC map was classified into FAO salinity categories and validated through independent field observations. This framework highlights the effectiveness of using freely available satellite data and cloud-based platforms like GEE for cost-effective soil salinity monitoring. This study provides a transferable methodology for precision agriculture, enabling informed land management and crop planning in salinity-affected regions.