Compound K (C-K) is a crucial pharmaceutical and cosmetic component because of disease prevention and skin anti-aging effects. For industrial application of this active compound, the protopanaxadiol (PPD)-type ginsenosides should be transformed to C-K. β-Glycosidase from Sulfolobus solfataricus has been reported as an efficient C-K-producing enzyme, using glycosylated PPD-type ginsenosides as substrates. β-Glycosidase from S. solfataricus can hydrolyze β-d-glucopyranoside in ginsenosides Rc, C-Mc1, and C-Mc, but not α-l-arabinofuranoside in these ginsenosides. To determine candidate residues involved in α-l-arabinofuranosidase activity, compound Mc (C-Mc) was docking to β-glycosidase from S. solfataricus in homology model and sequence was aligned with β-glycosidase from Pyrococcus furiosus that has α-l-arabinofuranosidase activity. A L213A variant β-glycosidase with increased α-l-arabinofuranosidase activity was selected by substitution of other amino acids for candidate residues. The increased α-l-arabinofuranosidase activity of the L213A variant was confirmed through the determination of substrate specificity, change in binding energy, transformation pathway, and C-K production from ginsenosides Rc and C-Mc. The L213A variant β-glycosidase catalyzed the conversion of Rc to Rd by hydrolyzing α-l-arabinofuranoside linked to Rc, whereas the wild-type β-glycosidase did not. The variant enzyme converted ginsenosides Rc and C-Mc into C-K with molar conversions of 97%, which were 1.5- and 2-fold higher, respectively, than those of the wild-type enzyme. Therefore, protein engineering is a useful tool for enhancing the hydrolytic activity on specific glycoside linked to ginsenosides.