We study a dynamic model of interactions between a firm and job applicants to understand long-term hiring outcomes. In each round, the firm decides which applicants to hire, where the firm's ability to evaluate each applicant is imperfect. Each applicant belongs to a group, and central to our model is the concept that the firm's evaluation ability is influenced by its past hiring decisions with the applicant's group. The dynamic nature of our model enables us to characterize the steady-state outcome of this process, which then reveals a mechanism of why disparities between groups may emerge. Specifically, we show that two groups of applicants with identical underlying skill distributions may converge to completely different long-term hiring outcomes, and this disparity is fueled by differences in the firm's initial ability to evaluate applicants of each group. This demonstrates that a minor initial disadvantage can amplify over time through a feedback loop, leading to persistent and exacerbating disparities even when firms make rational hiring decisions. Finally, we study the effectiveness of short-term interventions that aim to improve long-term hiring outcomes. Joint work with Ali Makhdoumi.