Detection of blood groups has to be very accurate and quick in clinical settings involving transfusions, surgeries, and emergency diagnostics. The research will implement a deep-learning-based automated blood group classification system with the latest YOLOv11x object detection model. It was trained on a richly diverse dataset comprised of 14,808 images captured under actual working conditions with superb results. Mean Average Precision (mAP@0.5) attained by the proposed YOLOv11x model is about 98.06% on the main test set and 98.45% on an unseen validity set; hence, robustness and generalizability have been tested and confirmed. In prediction task confidence scores were realized ranging from 0.75 up to 0.95 revealing high certainty of the model in classifying detections, that as compared to small-object detection in noisy low contrast images where challenges prevail due to reason another hybrid backbone CNN-Transformer was incorporated plus Multi augmentation like Brighticty Modulation Rotation Contrast Variation Noise injection. These techniques significantly improved model resilience toward light variations, occlusion, and color distribution. The system is implemented in Python and facilitated as a web app through FastAPI and Docker; hence, blood group detection can be done in real time without the need for any local installation. The entire model pipeline was trained on Google Colab using NVIDIA 3090 GPUs with transfer learning from COCO-pretrained weights of YOLOv5. This automated method reduces the risk of human error in blood typing, makes processing time faster, and is scalable for hospital deployments to rural clinics. The study validated the effectiveness of YOLOv11x medical image processing while paving the way for further intensification of accuracy through more visionary transformers besides dataset widening.