Steroidal compounds are some of the most prescribed medicines, being indicated for the treatment of a variety of conditions including inflammation, heart disease, and cancer. Synthetic approaches to functionalized steroids are important for generating steroidal agents for drug screening and development. However, chemical activation is challenging because of the predominance of inert, aliphatic C–H bonds in steroids. Here, we report the engineering of the stable, highly active bacterial cytochrome P450 enzyme P450BM3 (CYP102A1) from Bacillus megaterium for the mono- and dihydroxylation of androstenedione (AD), dehydroepiandrosterone (DHEA), and testosterone (TST). In order to design altered steroid binding orientations, we compared the structure of wild type P450BM3 with the steroid C19-demethylase CYP19A1 with AD bound within its active site and identified regions of the I helix and the β4 strand that blocked this binding orientation in P450BM3. Scanning glycine mutagenesis across 11 residues in these two regions led to steroid oxidation products not previously reported for P450BM3. Combining these glycine mutations in a second round of mutagenesis led to a small library of P450BM3 variants capable of selective (up to 97%) oxidation of AD, DHEA, and TST at the widest range of positions (C1, C2, C6, C7, C15, and C16) by a bacterial P450 enzyme. Computational docking of these steroids into molecular dynamics simulated structures of selective P450BM3 variants suggested crucial roles of glycine mutations in enabling different binding orientations from the wild type, including one that closely resembled that of AD in CYP19A1, while other mutations fine-tuned the product selectivity. This approach of designing mutations by taking inspiration from nature can be applied to other substrates and enzymes for the synthesis of natural products and their derivatives.