Tomatoes are among the most widely cultivated crops worldwide, but they remain especially vulnerable to insect pests that can drastically reduce both yield and quality. Traditional monitoring methods, which often rely on manual observation, are time-consuming, prone to human error, and difficult to scale in large agricultural settings. To address this challenge, we present a smart, embedded system designed to automatically detect and classify tomato pests in real time. Our solution focuses on three major pest species: Tuta absoluta, Bemisia tabaci, and Tetranychus urticae. Unlike previous works, this system integrates a two-stage architecture combining YOLOv11n for fast object detection and YOLOv11-cls for fine-grained classification, specifically optimized for low-power embedded deployment on platforms like the Jetson Nano and Raspberry Pi 5. The system achieved a detection performance of 71.2% mAP@0.5 and 42.3% mAP@0.5:0.95 (using the selected YOLOv1ln model) and a classification accuracy of 93.3%. With low memory usage and high inference speed, the system proves highly suitable for embedded deployment. This work offers a practical, low-cost solution to support farmers and researchers in pest management, contributing to more sustainable and data-driven agriculture. By enabling farmers to monitor pest populations more precisely, the system helps improve decision-making in pest control, offering a cost-effective, real-time solution to reduce reliance on excessive pesticide use.