Remembering people is at the core of many social and economic relationships. We present evidence of systematic biases in the way we remember people, based on two experiments. The first experiment is conducted in a real professional setting - academia. Participants of two academic conferences are asked to recall ‘who presented what’ a month after attending the conferences. The second experiment is a controlled version of the first. Participants are shown pictures of people, matched with the title of a paper. We exogenously vary the relative shares of women and non-white individuals. In both experiments, we find evidence that women and ethnic minorities are more likely to be remembered in settings where they are in a small minority. In contrast, they are more likely to be confused with each other when they are in larger fraction. These findings are in line with a theory of categorization. People with minority attributes appear to be “blended together.” We conjecture that these biases in remembering could have important implications for the formation of professional networks.