Accurate soil deficiency diagnosis and effective irrigation scheduling are the critical components of agricultural output, yet traditional methods find it difficult to adjust to changing environmental conditions. While standard soil testing is unable to identify nutrient deficits in real time, fixed irrigation plans result in the water mismanagement. The accuracy of classical machine learning models in agricultural decision-making is decreased by their limited adaptability and inadequate feature selection. In order to get of these restrictions, this study suggests a Boosting-based Prediction and Auto-Optimization Model that uses CatBoost for soil deficiency analysis and LightGBM for irrigation scheduling. Macronutrients and Micronutrients data from Tamil Nadu villages are included in the analysis of irrigation and soil deficiency analysis. In terms of soil qualities, the Water Requirement Index (WRI) divides irrigation zones into Low, Medium and High categories, while the Soil Deficiency Index (SDI) measures the nutrient imbalances. Achieving 96% accuracy in nutrient-based soil categorization, the CatBoost model effectively detects deficiencies and optimizes fertilizer recommendations. The LightGBM model, on the other hand, forecasts WRI with 99% accuracy using the K-Means clustering for irrigation zoning, guaranteeing accurate irrigation scheduling depending on soil nutrient levels. In contrast to traditional models, this method dynamically modifies soil management techniques and irrigation schedules to increase nutrient efficiency and water saving. This paper depicts those farmers can use this AI driven insights to make data-driven decisions that will improve the crop productivity and health.