Soil-transmitted helminths (STH), including hookworm, Ascaris lumbricoides, and Trichuris trichiura, impose significant health burdens in low- and middle-income tropical and subtropical regions, infecting over 1.5 billion people globally. Traditional diagnostic methods like the Kato-Katz technique are time consuming. This study introduces an innovative AI-driven system utilizing affordable 3D-printed adapters and smartphones to digitize Kato-Katz microscopy samples, capturing high-resolution images for subsequent analysis. These digitized images can be uploaded to a telemedicine platform for remote diagnosis and expert consultation. Central to our system is the development of a foundational AI model for parasite detection and classification. The model operates in two stages: First, an object detection algorithm identifies all parasites in the image, achieving a mean average precision (mAP) of 97.90% on the validation set using the YOLOv8 architecture. Second, a classification algorithm categorizes each detected parasite by species. The classification model is initially trained on a large, unannotated dataset of parasite images using a self-supervised learning (SSL) approach to learn domain-specific visual features, which are often missed while using generic pre-training datasets. Subsequently, it is fine-tuned on a labeled dataset, significantly improving performance. The model initialized with SSL on STH images achieved an F1 score of 91.70%, outperforming those initialized with random weights (F1 score of 55%) and those trained on DINO-Imagenet weights (F1 score of 53%). By integrating AI with low-cost digital imaging, our approach aims to revolutionize STH diagnosis in resource-constrained settings, aligning with the WHO's 2030 Roadmap for the elimination of neglected tropical diseases.