The analysis starts with an OLS model to explain salary variation using demographic and organizational factors. Diagnostic tests revealed issues such as influential observations, heteroscedasticity, non-normal residuals, and multicollinearity. To address these limitations, robust regression and regularized models (Ridge, Lasso, and Elastic Net) were applied. The results showed that regularization improved predictive performance and model stability, with Elastic Net providing the best balance between accuracy, robustness, and interpretability.