The key precursors for nylon synthesis, that is, 6‐aminocaproic acid (6‐ACA) and 1,6‐hexamethylenediamine (HMD), are produced from petroleum‐based feedstocks. A sustainable biocatalytic alternative method from bio‐based adipic acid has been demonstrated recently. However, the low efficiency and specificity of carboxylic acid reductases (CARs) used in the process hampers its further application. Herein, we describe a highly accurate protein structure prediction‐based virtual screening method for the discovery of new CARs, which relies on near attack conformation frequency and the Rosetta Energy Score. Through virtual screening and functional detection, five new CARs were selected, each with a broad substrate scope and the highest activities toward various di‐ and ω‐aminated carboxylic acids. Compared with the reported CARs, KiCAR was highly specific with regard to adipic acid without detectable activity to 6‐ACA, indicating a potential for 6‐ACA biosynthesis. In addition, MabCAR3 had a lower Km with regard to 6‐ACA than the previously validated CAR MAB4714, resulting in twice conversion in the enzymatic cascade synthesis of HMD. The present work highlights the use of structure‐based virtual screening for the rapid discovery of pertinent new biocatalysts.