Chronic Kidney Disease (CKD) has become a global health crisis, showing rapid increases across all age groups. Frequently termed a "silent killer" , CKD often evades detection in its early stages , necessitating complex diagno- -stic assessments.While machine learning (ML) models offer promising pathways for early detection , achieving high predictive accuracy with limited datasets remains a formidable challenge , often complicated by noise and overfitting. This study proposes an optimized framework utilizing advanced feature engineering and data positioning to enhance model reliability. We evaluated the performance of nine distinct machine learning models—including LightGBM, SVC, and Random Forest—using metrics such as accuracy, precision, recall, F1-score, and Mean Squared Error (MSE). Our analysis demonstrates that the LightGBM model achieved an individual accuracy of 90.06% with an MSE of 0.0994. Furthermore, by employing ensemble techniques, we achieved a peak accuracy of 93.07%. These results suggest that with rigorous algorithmic optimization and sophisticated data processing, it is possible to diagnose CKD with high precision even in data-constrained clinical environments. This research provides a scalable, reliable diagnostic tool, marking a significant advancement in medical informatics and early disease management.