How should gender discrimination and systemic disadvantage be addressed when more discriminatory and less generous students systematically sort into certain fields, courses, and instructors' sections? In this paper, we estimate measures of gender bias and evaluation generosity at the student level by examining the gap between how a student rates male and female instructors, controlling for professor fixed effects. Accounting for measurement error, we find significant variation in gender bias and generosity across students. Furthermore, we uncover that bias varies systematically by gender and field of study and that patterns of sorting are sufficiently large to place female faculty at a substantive disadvantage in some fields and male faculty at a disadvantage in others. Finally, we document that sexist attitudes are predictive of gender-based sorting and propose Empirical Bayes inspired measures of student-level bias to correct for instructor-specific advantages and disadvantages caused by sorting.