Deoxynivalenol (DON), the most common mycotoxin in the world, poses serious health hazards to both humans and animals and causes significant economic losses to the food and feed industry. An intriguing method for DON detoxification involves enzymatic cascade catalysis by pyrroloquinoline quinone-dependent alcohol dehydrogenase (DADH) and aldo-keto reductase (AKR13B3). However, the poor catalytic activity of AKR13B3 has limited its usefulness. Here, structure-guided steric hindrance engineering, computer-assisted protein engineering, and combinatorial mutagenesis were carried out to increase the catalytic efficiency of AKR13B3. The best mutant in this project was M28S/S65V, which outperformed wild-type AKR13B3 in terms of catalytic efficiency (kcat/Km) and specific activity by factors of 44.07 and 41.65, respectively. Kinetic parameter determination revealed that the enhanced catalytic efficiency of M28S/S65V toward 3-keto-DON was attributed to the decreased Km and increased kcat values. Moreover, the engineered enzyme was applied to degrade DON in contaminated corn steep liquor, for a 90.62% removal rate. Structure-based computational analysis provided insights into the enhanced catalytic efficiency of M28S/S65V, which was attributed to the enlarged substrate-binding pocket and reshaped pocket with a favorable hydrophilic attack distance. The results provide a reference for improving the catalytic activity of an enzyme toward non-native bulky substrates and pave the way for the development of enzymes as detoxification agents to mitigate DON in the food and feed industry.