The general goal of this doctoral thesis is the development of tools for the production and application of two glycosidic enzymes: alpha-L-arabinofuranosidase from the filamentous fungi Aspergillus niger (Abf) and beta-D-glucosidase from the yeast Candida molischiana (Bgl). These hydrolases are used to release sugars in biomass conversion processes and food industry, and to synthesise aminoglycosides, glycoconjugates and oligosaccharides, high-added value compounds for chemical and pharmaceutical industries. The genes coding for these enzymes have been expressed in the methylotrophic yeast Pichia pastoris, and the produced enzymes purified in order to characterise their biochemical properties. Their capacity to catalyse transglycosylation reactions in high yield has been validated. In view of production process assessment and monitorisation, a constraint-based model of P. pastoris metabolism has been devised and validated using possibilistic metabolic flux analysis, in order to evaluate model consistency and estimate biomass growth rate and intracellular flux distributions using only a few extracellular measured rates. The model has been extended to estimate recombinant protein productivity, and used to analyse different growth conditions of the recombinant strains overproducing Abf and Bgl enzymes. Finally, the recombinant enzymes have been immobilised on bimodal organosilicas of the UVM-7 family. The resulting biocatalysts have been biochemically and physicochemically characterised and evaluated in different applications of biotechnological interest.