Tomato crops are vulnerable to various disease that threaten food security and farmers livelihoods, especially in resource -limited regions like Tigray, Ethiopia. This study presents a hybrid deep learning model combining Resnet-50 and MobilenetV2 with Explainable AI (Grad-CAM) for automated tomato leaf disease classification. Using a benchmark dataset of 17,920 annotated images across 10 disease classes, Resnet-50 achieved training accuracy 99.25% and 94.5% validation accuracy, while MobilenetV2 reached training accuracy 94.6% and validation accuracy 89%, offering a trade-off between accuracy and deployment feasibility. Grad-cam visualization confirmed model focus on biologically relevant leaf regions, enhancing interpretability. The proposed system demonstrates potential for real-world deployment in precision agriculture and supports small holder farmers through early disease detection and informed crop management.