The affinity engineering is a key step to increase the efficacy of therapeutic monoclonal antibodies and yeast surface display is the most widely used and powerful affinity maturation approach, achieving picomolar binding affinities. In this study, we provide an optimization of the yeast surface display methodology, applied to the generation of potentially therapeutic high affinity antibodies targeting the immune checkpoint PD-L1. In this approach, we coupled a 10-cycle error-prone mutagenesis of heavy chain complementarity determining region 3 of an anti‐PD-L1 scFv, previously identified by phage display, with high-throughput sequencing, to generate scFv-yeast libraries with high mutant frequency and diversity. In addition, we set up a novel, faster and effective selection scheme by fluorescence-activated cell sorting, based on a fast drop of the antigen concentration between the first and the last selection cycles, unlike the gradual decrease typical of current selection protocols. In this way we isolated 6 enriched mutated scFv-yeast clones overall, showing an affinity improvement for soluble PD-L1 protein compared to the parental scFv. As a proof of the potency of the novel approach, we confirmed that the antibodies converted from all the mutated scFvs retained the affinity improvement. Remarkably, the best PD-L1 binder among them also bound with a higher affinity to PD-L1 expressed in its native conformation on human-activated lymphocytes, and it was able to stimulate lymphocyte proliferation in vitro more efficiently than its parental antibody. This optimized technology, besides the identification of a new potential checkpoint inhibitor, provides a tool for the quick isolation of high affinity binders.