The invisible nature of microbial life belies its staggering diversity and ubiquity. These unseen organisms impact the eco-evolutionary processes of the macro world at levels ranging from landscape-level nutrient cycling to directly impacting the fitness of individual microbiome hosts. Consequently, microbes and microbiomes are increasingly recognized for their potential applications in agriculture, biomanufacturing, and medicine; however, the identification of strains and engineering of consortia best-suited for a given application remains a challenge with nascent design principles. Where the application involves a host, host-mediated microbiome selection can be used to identify simplified consortia of microbes that elicit desired host phenotypes.In brief, host-mediated microbiome selection applies a selection pressure to a population of hosts with the expectation that variation in host phenotype results from variation in their respective microbiomes. A subset of microbiomes—typically from the best-performing hosts—can then be selected and used to inoculate a subsequent cohort of hosts and the process repeated to iteratively derive host-associated microbiomes with the capacity to elicit a host phenotype of interest. While this form of artificial selection performed at the community level is not new, there are relatively few examples of its use in the literature. In the following chapters I will describe a number of experiments and methods development that led to the successful use of host-mediated microbiome selection to improve drought tolerance in rice. Chapters 1 and 2 provide a general introduction to plant-microbiome interactions and describe the methods I developed to enable non-destructive, high-throughput phenotyping of droughted plants. Chapter 3 characterizes the diverse set of field soils collected as candidate source communities for initial microbiome inoculation and details a number of preliminary experiments that helped to optimize rice growth conditions and inoculation protocols. Finally, Chapter 4 reports the results from multiple host-mediated microbiome selection experiments where each experiment used a different field-soil community as its microbiome source. Broadly speaking, the last several years spent planning and performing host-mediated selection experiments have led me to consider the following as important guide rails for their design: 1) Minimizing “noise” in host phenotypes—variation due to environmental fluctuations, cross-contamination, etc.—should be a priority; the high dimensionality, and in most cases compositionality, pose enough challenge to the interpretation of microbiome data. 2) The provenance of microbiomes matter; though a source of inoculum is as easily obtained as scooping up a handful of soil, the landscape is rife with gradients along which there may be microbial communities naturally predisposed to survive the conditions into which they will be inoculated or to carry functions relevant to the phenotype of interest. 3) Rapid cycling of selection generations selects for fast-growing microbes and/or those with efficient dispersal and colonization habits; this subset of the initial community may be limited in its ability to impact host phenotypes. 4) The balance between the number of independent selection lines and replicates per line should be carefully considered; replicate lines are necessary to demonstrate that final microbiomes were arrived at through more than a random walk, but relatively high levels of replication within a line are required to identify specific microbial taxa associated with a phenotype. What follows are the details of experiments that led me to these conclusions which I hope can provide useful insights to others attempting to employ host-mediated microbiome selection.