Designing efficient enzymes is a formidable challenge at the forefront of modern biocatalysis. Here, we review recent developments in the field and illustrate how the interplay between computational design and advanced protein engineering has given rise to enzymes with diverse activities. Natural proteins have been re-engineered computationally to embed designed catalytic sites, affording active catalysts that can be optimized through laboratory evolution to enhance efficiency and selectivity. Computational design tools can reliably generate stable de novo proteins with shapes and backbone geometries beyond those found in nature, which can serve as idealized templates for hosting catalytic sites. Genetic code reprogramming methods have been used to introduce additional functional elements into protein active sites to expand the range of chemistries accessible with designer enzymes. Finally, the recent emergence of powerful protein design tools based on deep learning promises to have a transformative impact on the field by greatly increasing the design speed and model accuracy. By bringing together the latest computational and experimental tools for enzyme design, we are optimistic that the ambition of reliably building useful biocatalysts from scratch is within reach.