Geometric deep learning is a new theoretical concept in AI that integrates conventional deep neural networks into non-Euclidean space. It combines symmetry and invariance in manifold structures and gauge-equivariant in theoretical physics, making it promising for edge computing and tiny machine learning in sensing applications. Researchers hope to use this technology to understand and predict the development and distribution of COVID-19, ultimately improving public health systems. Manuscripts on this topic can be submitted for publication.