The pathogenicity and transmissibility of a virus are strongly governed by spatial patterns of viral growth and spread across the internal tissue landscape of the host individual. These ‘within-host dynamics’ can differ widely among pathogens, and even among individuals infected with the same pathogen. Epidemiologists and microbiologists have long understood that many factors contribute to this observed heterogeneity, including the dose and route of exposure, host demographic and health factors, the cellular properties of different tissues (e.g., receptor expression), viral sensitivity to within-host abiotic conditions (e.g., temperature, pH), and physical connectivity between tissues (e.g., via blood). However, the relative effects of these governing processes on realized infection patterns inside hosts are difficult to disentangle and remain poorly understood, largely due to sampling constraints and data limitations. For humans, natural exposure events are inherently unobservable, and available data can be biased towards individuals with more severe disease that seek treatment. Animal challenge experiments can control the relevant dimensions, but they either have limited clinical relevance due to crucial physiological differences from humans (e.g., for small animal models), or their small sample sizes cannot support robust statistical inferences under the standard approach of analyzing data only within the study that generated them (e.g., for non-human primate models).Modern, data-driven computational models offer a powerful but underutilized toolkit to overcome these observational and analytical limitations. By using Bayesian statistical approaches to integrate quantitative modeling techniques with experimental data, it is possible to extract underlying patterns and putative mechanisms from limited empirical observations and to enhance these insights by jointly analyzing disparate datasets. In this dissertation, I develop and apply such methods to characterize the effects of exposure route, exposure dose, tissue connectivity, tissue traits, and host demographic factors on within-host SARS-CoV-2 dynamics. These analyses are supported by a large database of 107 studies that I have constructed, which includes 22,183 viral measurements from 721 non-human primates that were experimentally challenged with SARS-CoV-2 by various routes and with various doses. In chapter 1, I address the fundamental question of when (if ever) RT-qPCR measurements of viral RNA load can reliably indicate the presence of infectious virus in a sampled tissue. This work constitutes the largest analysis of this question using in vivo infection data from individuals with known exposure conditions, and it lays crucial groundwork for the application of my customized statistical approach to public health contexts. I demonstrate that total RNA measurements can indeed predict culture positivity with a remarkable 85% accuracy on out-of-sample data as long as predictions also account for other factors, including exposure conditions, host demographics, and assay protocols. In chapter 2, I conduct the first large-scale quantitative analysis of respiratory virus challenge experiments in non-human primates to characterize the relative impacts of exposure route, exposure dose, age, sex, and species on within-host dynamics. I show that exposure route more strongly modulates the probability, onset, peak, and conclusion of SARS-CoV-2 infection across the respiratory and gastrointestinal tracts than exposure dose or demographic factors. I also show that infection patterns following aerosol inhalation are clearly distinct from any other exposure route, including intranasal or combined intranasal/intratracheal inoculation. This work provides the most comprehensive and quantitative evidence to date that exposure conditions shape infection patterns inside hosts, in ways that affect disease risk and shedding potential. In chapter 3, I develop a novel modeling framework that formally investigates how tissue traits (e.g., receptor expression, protease availability) and connectivity structure interact to determine spatiotemporal infection patterns inside hosts, which I fit to the data from nine challenge studies. This model shows that SARS-CoV-2 infection patterns across the respiratory and gastrointestinal tracts are shaped by high rates of within-host viral dissemination. I also show that infections are overall more successful in the nose and throat than in the lung and lower GI, which is consistent with an estimated increase in the local infection rates at the lower ambient temperature of the upper respiratory tract. Together, these chapters demonstrate that meta-analysis of the data from in vivo challenge experiments can overcome the difficulties arising from limited sample sizes in crucial but costly animal models and that they can yield robust insights beyond those attainable from individual studies, all while reducing overall animal use in infectious disease research. This dissertation focuses on characterizing SARS-CoV-2 infections in non-human primates, but the methods developed here can be readily adapted to any other pathogen-host system, and they present generalizable, quantitative approaches to answer questions at the frontier of virology.