7β-Hydroxysteroid dehydrogenases (7β-HSDHs) have attracted increasing attention due to their crucial roles in the biosynthesis of ursodeoxycholic acid (UDCA). However, most published 7β-HSDHs are strictly NADPH-dependent oxidoreductases with poor activity and low productivity. Compared with NADPH, NADH is more stable and cheaper, making it the more popular cofactor for industrial applications of dehydrogenases. Herein, by using a sequence and structure-guided genome mining approach based on the structural information of conserved cofactor-binding motifs, we uncovered a novel NADH-dependent 7β-HSDH (Cle7β-HSDH). The Cle7β-HSDH was overexpressed, purified, and characterized. It exhibited high specific activity (9.6 U/mg), good pH stability and thermostability, significant methanol tolerance, and showed excellent catalytic efficiencies (k_cat/K_m) towards 7-oxo-lithocholic acid (7-oxo-LCA) and NADH (70.8 mM^-1s⁻¹ and 31.8 mM^-1s⁻¹, respectively). Molecular docking and mutational analyses revealed that Asp42 could play a considerable role in NADH binding and recognition. Coupling with a glucose dehydrogenase for NADH regeneration, up to 20 mM 7-oxo-LCA could be completely transformed to UDCA within 90 min by Cle7β-HSDH. This study provides an efficient approach for mining promising enzymes from genomic databases for cost-effective biotechnological applications.