Abstract Background Genetic markers are pivotal to modern genomics research; however, discovery and genotyping of molecular markers in oat has been hindered by the size and complexity of the genome, and by a scarcity of sequence data. The purpose of this study was to generate oat expressed sequence tag (EST) information, develop a bioinformatics pipeline for SNP discovery, and establish a method for rapid, cost-effective, and straightforward genotyping of SNP markers in complex polyploid genomes such as oat. Results Based on cDNA libraries of four cultivated oat genotypes, approximately 127,000 contigs were assembled from approximately one million Roche 454 sequence reads. Contigs were filtered through a novel bioinformatics pipeline to eliminate ambiguous polymorphism caused by subgenome homology, and 96 in silico SNPs were selected from 9,448 candidate loci for validation using high-resolution melting (HRM) analysis. Of these, 52 (54%) were polymorphic between parents of the Ogle1040 × TAM O-301 (OT) mapping population, with 48 segregating as single Mendelian loci, and 44 being placed on the existing OT linkage map. Ogle and TAM amplicons from 12 primers were sequenced for SNP validation, revealing complex polymorphism in seven amplicons but general sequence conservation within SNP loci. Whole-amplicon interrogation with HRM revealed insertions, deletions, and heterozygotes in secondary oat germplasm pools, generating multiple alleles at some primer targets. To validate marker utility, 36 SNP assays were used to evaluate the genetic diversity of 34 diverse oat genotypes. Dendrogram clusters corresponded generally to known genome composition and genetic ancestry. Conclusions The high-throughput SNP discovery pipeline presented here is a rapid and effective method for identification of polymorphic SNP alleles in the oat genome. The current-generation HRM system is a simple and highly-informative platform for SNP genotyping. These techniques provide a model for SNP discovery and genotyping in other species with complex and poorly-characterized genomes.