Stochasticity in gene-expression can be detrimental to fitness by randomizing developmental pathways, but can be advantageous when heterogeneous response is required. Using a combination of tools derived from biochemistry, microfluidics, microscopy, and engineering, we are trying to identify and characterize the broad principles underlying the origin and role of stochasticity in bacterial cell-fates and fitness. To analyze such processes in detail, we need time-lapse data of single cells, with high-sensitivity and deep-sampling that allow us to detect low copy-number molecules and rare-events, which are often the important drivers for cell-fate. We have developed a novel phenotyping platform that enables us to monitor the dynamics of physiology and gene-expression of >105 cells in parallel with tightly controlled growth-conditions, in dense culture or eco-systems. We can also harness the throughput to phenotype cells from thousands of different strains in parallel, from genomic library or clinical isolates, using a novel self-erasable barcoding system. The barcodes also enable us to run multi-strain competition experiment with randomized position in the microfluidic device, allowing us to resolve fitness differences down to 0.01%. I will present a few stories illustrating how the ability to track phenotypes over time, observe exceedingly rare events, and ultraprecise measurement of fitness differences has allowed us to uncover several novel principles about fate-choice consequences as single microbes try to eliminate or exploit their internal stochasticity in the face of an ever-fluctuating external environment.