In the context of increasing demand for food quality inspection and automation in agriculture, this paper presents the design and development of an intelligent tomato sorting system that integrates artificial intelligence (AI) and robotic control. The system combines image processing techniques with a convolutional neural network (CNN) model to classify tomatoes into three categories: ripe, unripe/deformed, and damaged. RGB images of tomatoes are captured in real time via a webcam and processed to extract visual features such as color, shape, and surface condition. The CNN model is trained on a custom dataset and deployed on a Raspberry Pi platform to perform on-device inference. The classification results are transmitted via serial communication to a 3-degree-of-freedom (3-DOF) robotic arm controlled by an Arduino. The robotic arm is responsible for picking and sorting the tomatoes into designated locations based on the predicted results. The control system includes a conveyor belt, infrared sensors for position detection, and a gripping mechanism utilizing stepper motors and limit switches. The proposed system demonstrates high classification accuracy and real-time automation capability, highlighting its potential for smart agriculture and post-harvest processing applications. In the future, the system could be extended to recognize other types of agricultural products and improve adaptability under diverse lighting conditions.