Coronary Artery Disease (CAD) is the main reason of cardiovascular death. Timely and accurate prediction is very much needed to enhance patient outcomes. Computed Tomography Angiography (CTA) is generally a Non-intrusive imaging technique employed to evaluate CAD; however, manual analysis consumes more time and is prone to observer variability. Our research proposes a framework based on deep learning to automate the detection of CAD in CTA images employing Convolutional Neural Networks (CNNs) and attention mechanisms for robust Feature Extraction as well as Classification. The pipeline integrates sophisticated pre-pre-processing methods for example vessel segmentation, contrast enhancement, as well as Noise Reduction to enhance the quality of the image. Our proposed Model is trained and tested using high-scale Annotated Datasets to obtain more accuracy, sensitivity, and specificity than conventional methods. The outcomes show how deep learning improves CAD diagnosis, minimizing diagnostic delays and inter-observer variability. Our future study will explore the mixture of multimodal imaging data, explainable AI methods, and conical validation to boost model interpretability and real-world applications.