As an increasing amount of water scarcity becomes the harsh reality, irrigation optimization becomes unavoidable in today’s farming. In this paper, we introduce an end-to-end smart irrigation system with an IoT installation and a hybrid deep learning framework supporting autonomous watering. Our extensible system integrates real-time sensor measurements from environmental sensors (e.g., soil moisture, temperature) via an IoT gateway, an efficient preprocessing pipeline, and automated watering control. Its fundamental technological innovation is a Transformer–GRU hybrid architecture for high-accuracy reference evapotranspiration (ET \(_0\) ) prediction. Through the use of the Transformer’s capacity to capture long-range dependency with the GRU’s advantage in sequential dynamics, our architecture achieves outstanding predictive efficacy ( \(R^2 = 0.995\) ). These spatiotemporally specific ET \(_0\) forecasts are continuously monitored in real time to create needs-oriented irrigation schedules, which are applied by the controllers of the farm’s system. The scalable architecture represents an integral remedy for precision farming, with considerable reduction in the wastage of irrigation water, thereby closing the gap between sophisticated AI prediction and practical, automated IoT exploitation at the field level.