This research is focused on developing techniques for subgraph representation learning. This task was prompted by a large financial dataset provided by an industry partner for the purpose of detecting illicit behavior in transaction networks. This is a challenging problem since it requires multitudes of relationships inside and outside the subgraph to be captured in a single vector. This research aims to build a model that is able to achieve state of the art performance on subgraph representation tasks, and more specifically, able to perform well on the financial dataset. Model development is currently based on a pretrain then prompt framework, and the bulk of the research will be focused on developing specific subgraph prompting methods, which have yet to be explored in the community.

I am participating in the SuperUROP program to better prepare myself for graduate school. I have done UROPs in the past, but often felt that I needed a stronger commitment to produce meaningful results. I hope to contribute significantly to the field of graph learning, and in that pursuit, become a better researcher and hacker. What excites me most about my project is its potential to be implemented in real-life financial security systems.