Weeds are one of the most important factors in crop production. Farmers often experience poor agricultural yields as a result of weeds. In India, traditional weed control relies on manual, mechanical, or chemical methods, which are labor-intensive, costly, and often inefficient. We present a low-cost, AI-powered prototype using computer vision to autonomously detect weeds in real time and simulates targeted herbicide application. Our proposed system comprises a Raspberry Pi attached to a camera module, to capture field imagery. An object detection model hosted locally on a server is then used to identify and localize the weeds, and an ESP32-controlled system activates LEDs representing herbicide nozzles based on predicted coordinates. This simulation demonstrates how precise, localized spraying can reduce chemical use and labor dependence. Our system achieved high accuracy in weed detection, with a mAP@0.50 of 96.9%, F1 score of 94.47%, and recall of 95.22%. Unlike bulky or cloud-dependent solutions, it runs efficiently on lightweight edge devices. Designed for rural deployment, it offers a scalable and sustainable AI-driven approach to weed management.