During infection, microbial pathogens encounter phagocytic cells of the host innate immune system, such as macrophages and neutrophils. These encounters typically lead to uptake and killing of the bacteria by the host cell or, conversely, parasitization of the host cell by the bacteria. Microscopic analysis of these host-microbe interactions is complicated by the fact that phagocytic cells are motile and tend to stray out of the field of view. Most host-pathogen studies have therefore been undertaken at the population level. However, the success of many pathogens likely lies in their ability to generate heterogeneity within the population, facilitating adaptation to every host micro-environment and to environmental perturbations (such as application of an antibiotic). We believe that studying host-pathogen interactions at the single-cell level can reveal not only new mechanisms of pathogenesis but also why current treatments are not always successful. This thesis presents a novel microfluidic device for the long-term co-culturing of phagocytic cells and bacteria within spatially confining microchambers. Each device is patterned with thousands of microchambers, permitting high-throughput fluorescence timelapse microscopy of host-microbe interactions at single-cell resolution. We demonstrate the utility of this approach by co-culturing an established host-cell model, Dictyostelium discoideum, with the extracellular pathogen Klebsiella pneumoniae. We observed that K. pneumoniae is readily taken up and killed by D. discoideum but some events can take ten times as long as others. The capsule, a polysaccharide layer present at the surface at the bacterial cell wall, reduces the phagocytosis efficiency by 1.5 fold, but does not influence the killing time. The opportunistic human pathogen Mycobacterium marinum was then used to study intracellular bacterial virulence. As expected, bacteria could infect host cells, replicate inside, and eventually lyse them. However, surprisingly, in the majority of the infections, the bacteria were actually released from the host, with no apparent harm to either the host or the bacteria. Less frequently, bacteria were even seen being killed. The timing of these events was highly heterogeneous, ranging from a couple of hours to several days. In summary, the tools presented in this thesis provide a new and simple approach for following individual host-microbe interactions at high spatiotemporal resolution from the onset to the end of infections. This work unveiled that the timing and outcome of host-microbe interactions are unexpectedly heterogeneous, which may have important consequences for the bacterial pathogenesis and ultimately for the development of future treatment of microbial infections.