Data-driven decision-making has become an essential part of modern life by virtue of the rapid growth in data, the massive improvements in computing power, and great progress in academic research. The range of techniques used fall broadly on the spectrum that varies from model-based to applied, depending on the problem complexity and data availability. This thesis studies three settings that span the modeling spectrum in the contexts of digital agriculture, cell reprogramming, and pandemic policymaking. First, we investigate the problem of learning good farming practices in the framework of multi-armed bandits with expert advice. We extend the setting from finitely many experts to any countably infinite set and provide algorithms that are provably optimal. Second, we explore optimizing perturbations for cell reprogramming in batched experiments. Building upon multi-armed bandit algorithms, we propose an active learning approach that integrates deep learning and biology-based analysis. We numerically demonstrate the success of our method on gene expression data. Finally, we model the impacts of nonpharmaceutical interventions during the coronavirus disease 2019 (COVID-19) pandemic. We develop an agent-based model in order to overcome the limitations of observational data. We show that the trade-off between COVID-19 deaths and deaths of despair, dependent on the lockdown level, only exists in the socioeconomically disadvantaged population. Our model establishes effective measures for reducing disparities during the pandemic.Ph.D.