Nature has utilized the power of evolution to create intricate biosynthetic pathways for the synthesis of various classes of natural products. Terpenes, the largest group of natural products, are primarily derived from only two substrates. Another large class, the polyketides, are known for their structural diversity and their many clinical uses. Type I polyketides, like the antibiotic Erythromycin A, are constructed via multi-megadalton polyketide synthase (PKS) ‘assembly lines’. The modular nature of these pathways has led to decades of interest in harnessing these enzymes for regioselective alterations of privileged scaffolds not easily accessible via synthetic or semisynthetic chemistry. Of particular interest for enzyme engineers, the acyltransferase (AT) domain in PKSs is responsible for the selection of the malonyl-CoA-derived extender unit substrate for incorporation by a given module. Despite rarely incorporating substrates beyond malonyl-CoA or methylmalonyl-CoA naturally, ATs arguably can provide the most added diversity to the polyketide natural product family through the selection of rare or non-natural malonyl-CoA derivatives. By shifting the selectivity of the AT, polyketide products can be diversified at every other carbon, providing chemical handles for new interactions and further semi-synthetic derivatization. To obtain a better understanding of AT selectivity, movements, and mutations, molecular dynamics simulations were performed (Chapter 2). Insight from the simulations led to identification of novel residues that contribute to AT selectivity. Successful mutations were used to inform swaps of AT structural motifs, leading to substrate selectivity shifts of greater than 300- fold while accounting for replacement of less than 3% of the AT’s sequence. During these in vitro studies, the ketoreductase (KR) domain was identified as a bottleneck for non-natural extender unit incorporation for the first time (Chapter 3). This lack of promiscuity by the AT, KR and other domains were explored more fully using the final two modules of the pikromycin PKS, PikAIII and PikAIV. The native and engineered promiscuities of both modules were compared, and AT mutagenesis enabled the first incorporation of two non-natural extenders into a full-length polyketide product (Chapter 4). To provide a robust source of natural and non-natural polyketide and terpene extender units, a directed evolution tool capable of screening or selecting for the best strain or enzyme is needed. Towards this goal, a transcription factor-based biosensor was refactored, optimized, and characterized with malonyl-CoA and the non-native methylmalonyl-CoA to provide extenders for new polyketide chassis (Chapter 5). To apply this technology towards the directed evolution of terpenes, a repressor protein recognizing linear monoterpenes was engineered and used as a test case for the development of more effective genetically-encoded biosensors (Chapter 6). Together, the described combined approaches of in silico, in vitro, and in vivo evolution of natural product biosynthetic enzymes provide a foundation of new knowledge and strategies for future natural product pathway engineering.