Ribonucleotide reductases (RNRs) are essential enzymes of life that produce the building blocks of DNA (2′-deoxyribonucleotides) by conversion from the building blocks of RNA (ribonucleotides). Despite the plethora of knowledge obtained for RNRs over the decades, many questions remain unknown due to the complexity and diversity of these enzymes. In this thesis, we describe progress towards further understanding the evolution and turnover of RNRs. Using phylogenetic and bioinformatic analyses, we report systematic investigation of the evolution of the RNR family, which revealed the presence of a novel ancestral clade as well as insights into the diversification of extensions and insertions around the RNR catalytic barrel. We then investigated the structural dynamics involved in class I RNR turnover using single particle cryo-electron microscopy (cryo-EM). By imaging the enzyme under near-native catalytic conditions, we obtained conformational landscapes of a class I RNR from cryo-EM datasets through advanced computational analyses, which revealed surprising asymmetry maintained across its turnover cycle. Finally, we report further efforts to characterize the allosteric sites known as the I- and M-sites, which evolved in the truncated ATP-cone domain of Bacillus subtilis class Ib RNR. We show direct evidence for the upregulation effect of guanine nucleotides at the M-site with activity assays, as well as bioinformatic analysis on the co-evolution of residues around the I-site and M-site. In addition to the study of RNRs, we also report the development of a tool for high-accuracy cryo-EM micrograph filtering based on convolutional neural networks. Overall, this thesis provides significant advances to the understanding of RNRs as well as a novel tool for cryo-EM processing.