Weed infestation remains a critical challenge to crop productivity and livelihood security in Indian agriculture, particularly in regions like Maharashtra, where yield losses of up to 80% occurs due to competition from weeds in crops such as soybean and cotton. This paper proposes a real-time inspired and low-cost weed detection system by combining ESP32-CAM module with YOLOv8 deep learning model for intelligent weed monitoring across large farmlands. The ESP32-CAM is used as a miniature IoT vision node which takes live field images and sends them over Wi-Fi to a detection system implemented with Python using the YOLOv8s model. The model was trained on a Roboflow dataset of more than 10,000 labeled images of crops and weeds and achieved mAP 50 = 0.85 with precision = 91.5% and recall = 85.6% for weed detection. The performance of the system is demonstrated in real-time inference with robust detection under varying illumination conditions. This work integrates embedded IoT vision systems with state-of-the-art deep learning methodologies for sustainable data-driven and automatic weed management in precision smart farming.