Explore ( https://about.instagram.com/features/search-and-explore ) is a discovery surface on Instagram ( https://www.instagram.com ) which sources content from across the platform based on a variety of factors such as accounts followed, and photos and videos liked. To enable discovery for a user, an aspect of this system is finding posts which are related to the content that a user has engaged with in the past. In this paper, we propose an approach to use session based signals to identify related posts. We create a graph with nodes as posts and session-bounded interactions as edge weight between these nodes. We then extract nearest neighbors for each node from this graph by means of Personalized PageRank. We test this approach in production at Instagram Explore as a candidate generation method for recommendation and observe a significant increase in user engagement, daily active users and user sessions in an AB test.