This study presents a novel AI-assisted multispectral infrared vision framework integrated into a robotic telerounding platform for postoperative monitoring. Leveraging convolutional neural networks (CNNs) and the proprietary generative AI model Aillumi, the system performs automated detection, classification, and documentation of thermos vascular abnormalities in surgical patients. Aillumi, trained on a curated dataset exceeding one million expert-annotated thermal images over a 10-year period, is capable of generating textual clinical reports from infrared image inputs with high precision. The robotic system incorporates a multispectral scanner that captures RGB and long-wave infrared (LWIR) imagery, enabling real-time perfusion analysis, wound monitoring, and early detection of complications such as deep vein thrombosis, infection, and vascular insufficiency. CNN architectures including InceptionV3, VGG16, and Xception were trained to identify abnormal infrared signatures using a preprocessing pipeline consisting of data augmentation, normalization, and ROI segmentation. Our results demonstrate high classification performance, with AUC values exceeding 0.98 for leading models. The generative output from Aillumi showed strong correlation with expert physician assessments, suggesting feasibility for autonomous postoperative documentation. This work underscores the role of integrated AI-robotic systems in advancing remote precision medicine and reducing hospital readmissions, with demonstrated potential in telemedicine and high-throughput surgical recovery environments, and future feasibility for rural healthcare deployments pending dedicated validation.