Nail deformities are precious but sometimes neglected pointers of systemic diseases like cancer, cardiovascular disease, and autoimmune diseases. Though clinically important, nail diseases have been under investigated in the field of artificial intelligence (AI). This paper introduces NailNet, a deep learning system inspired by the YOLOv12 architecture, capable of real-time, multi-class classification between six categories of nail conditions: Acral Lentiginous Melanoma, Clubbing, Pitting, Onychogryphosis, Blue Finger, and Healthy Nails. The model is trained on a precisely selected and balanced 3,200 high-quality image dataset, using sophisticated preprocessing, data augmentation, and transfer learning. NailNet performs well, with 92.4% Top-1 accuracy and 98.9% Top-5 accuracy. As opposed to standard CNN-based classifiers, NailNet has faster inference and is designed to be deployed on mobile phones, adequate for use in teledermatology, especially in resource-limited environments. The research focuses on the capability of real-time deep learning models in enabling early and convenient diagnosis of nail diseases worldwide.