While laboratory evidence has documented that men and women differentially perceive their abilities with respect to gender-conforming tasks, there is scant evidence about how behavior in these controlled environments maps to the real world and if targeted interventions can mold these gender differences. We make progress along both fronts by studying the effects of a gender stereotype training program on job preferences among high school seniors that choose between internships that are analytically-oriented versus service-oriented. The intervention, which will be implemented in 250 Peruvian high schools, will provide students with cognitive-emotional content and actionable strategies to counteract gender stereotypes that distort their perception of their own abilities as gender-specific over an extended semester-long in-person course. By collecting rich survey data, we can quantify baseline and endline gender differences in perceived abilities with respect to gender-conforming tasks, whose malleability we aim to assess. Further linking the intervention to a job choice model enables us to estimate the moderating impact of objective and self-perceived differences in women's and men's abilities on the selection of gender-conforming jobs, documenting the implications of these differences on real-world outcomes. The findings provide insights into how future student-based policies can reduce gender gaps in the labor market.