<p>This study presents an innovative approach to soil sampling and crop recommendation using unmanned aerial vehicles (UAVs) equipped with nitrogen-phosphorus-potassium (NPK) sensors, particularly targeting the riverbank regions of Punjab, India. Traditional soil sampling methods fall short in these ecologically diverse and physically inaccessible terrains. Leveraging the DJI Air 2S UAV and IoT-enabled NPK sensors, this research introduces a novel dataset encompassing essential agricultural parameters such as nitrogen, phosphorus, potassium, pH, temperature, and moisture. The data is integrated into a cloud-hosted crop recommendation platform enhanced with a novel hybrid machine learning approach, the PCA-based Stacked Ensemble Model (PSEM), which combines Support Vector Classifier, Random Forest, and Logistic Regression. The proposed model achieved superior classification performance with 84.21% accuracy and 87.51% precision, outperforming other standard machine learning algorithms. Crop recommendations tailored to specific riverine districts such as Tarn Taran, Kapurthala, Rupnagar, and Pathankot are also provided. The system is deployed on Amazon Web Services (AWS), enabling real-time access and scalability, and integrates SMS-based alerts to empower farmers with timely and data-driven crop and fertilizer suggestions. This comprehensive framework underscores the transformative potential of combining UAVs, cloud computing, and machine learning for sustainable and precision agriculture in remote and ecologically sensitive regions. Future enhancements of this system will explore real-time satellite data and deep learning models such as CNN-LSTM hybrids for improved temporal and spatial prediction capabilities.</p>