Brazil has experienced democratic backsliding in recent years, a process often blamed on rising distrust of a political class marred by corruption scandals. This phenomenon is also present in several other nations in both the developing and the developed world. In response, several non-profit organizations have emerged with training programs designed to select and train a new class of political leaders. But whether this response by civil society can succeed in renewing the political class remains uninvestigated. In this paper, we evaluate the largest non-profit, non-partisan training program for aspiring politicians in Brazil. Using a judge-specific regression discontinuity design, we estimate the effects of the program on aspirants deciding to run and, conditional on running, on their electoral performance during the 2020 local elections. We find that the program increased the likelihood that the aspirant ran for office, and that it led to modest improvements in their campaign funds and vote shares. We do not, however, find any evidence that these aspirants are more likely to get elected suggesting that the program has yet to have any meaningful impact on Brazil’s political selection. Our estimates inform program targets by identifying how large an effect on vote shares and finance improvements would be required for the program to impact political selection.