The genetic architecture of adaptation is fundamental to understanding the mechanisms and constraints governing diversification. However, most case studies focus on loss of complex traits or parallel speciation in similar environments. It is still unclear how the genetic architecture of these local adaptive processes compares to the architecture of evolutionary transitions contributing to morphological and ecological novelty. Here we identify quantitative trait loci (QTL) between two trophic specialists in an excellent case study for examining the origins of ecological novelty: a sympatric radiation of pupfishes endemic to San Salvador Island, Bahamas containing a large-jawed scale-eater and a short-jawed molluscivore with a skeletal nasal protrusion. These specialized niches and trophic traits are unique among over 2,000 related species. Measurements of the fitness landscape on San Salvador demonstrate multiple fitness peaks and a larger fitness valley isolating the scale-eater from the putative ancestral intermediate phenotype of the generalist, suggesting that more large-effect QTL should contribute to its unique phenotype. We evaluated this prediction using an F2 intercross between these specialists. We present the first linkage map for pupfishes and detect significant QTL for sex and eight skeletal traits. Larger-effect QTL contributed more to enlarged scale-eater jaws than the molluscivore nasal protrusion, consistent with predictions from the adaptive landscape. The microevolutionary genetic architecture of larger-effect QTL for oral jaws parallels the exceptional diversification rates of oral jaws within the San Salvador radiation observed over macroevolutionary timescales and may have facilitated exceptional trophic novelty in this system.