COVID-19 confronts us with menacing extremes and overwhelming uncertainties with which we struggle to grapple. Among these challenges is a fundamental incoherency of the data life-cycle approaches and connections between different scales, methods, and models critical in combating the pandemic. In this presentation I discuss how we might transition between models that vary in space and time, from individual to community, from county to city, and from the moment an infectious molecule first enters our bodies, to disease development, to community transmission. To that end, we synthesize ideas taken from the science of Public Health, Artificial Intelligence, Data Science, Computational and Combinatorial Geometry, Applied Mathematics, and Statistics and Network Science. Specifically, we compose a chain of Graphical Models (GMs) capable of predicting infection rates among differing communities using publicly available data, with the broader objective of computing the likelihood for a particular community in a city to transition from the initial percentage of infected inhabitants to a given number within a week, two weeks, or a month. We calibrate, train, and validate our pandemic-based GMs with Agent-Based Models (ABMs), the current state-of-the-art in pandemic modeling; the former (ABMs) provide greater resolution but the latter (GMs) are the fastest. The efficiency of GMs graduates modeling of the data lifecycle to the next level where we can not only quantitatively estimate spread but also build in the future methods and algorithms to mitigate the pandemic.
