Leveraging the diverse biochemistry of microorganisms, engineered living materials (ELMs) offer adaptive and sustainable alternatives to synthetic counterparts. Wild-type microorganisms used in ELMs must have a metabolic activity that matches an engineered need in a specific context. If this is not the case, they can be programmed genetically to achieve the desired functionalities. While synthetic biology tools advance, the rational design of ELMs relies on existing genotype-phenotype knowledge. Here, we introduce a high-throughput directed evolution platform to enhance microorganism fitness towards targeted phenotypes and integrate them into complex materials. Using Komagataeibacter sucrofermentans as a cellulose-producing model, our droplet-based microfluidic approach yields a few cellulose overproducers from 40'000 mutants. Sequencing of the native and evolved strains reveals an unanticipated link between cellulose formation and a protease gene, demonstrating the platform's ability to uncover genotype-phenotype connections. This novel strategy addresses challenges in enhancing microorganisms for specific applications and in discovering new genotype-phenotype links, marking a promising step toward the next generation of engineered living materials.