The Jackson School's Academic Workshop Series will host a presentation from Huifeng Su, a PhD student at the Yale School of Management, "Reuniting Forcibly Separated Families Through Shared Memories with Machine Learning." The Student Academic Workshop Series at the Jackson School of Global Affairs is a bi-weekly forum in which advanced undergraduate and graduate students can present their research projects and receive constructive feedback from their peers.

Over 100,000 victims rely on self-reported, semi-structured clues on humanitarian service platforms to search for their missing parent or child. Yet, these searches face vast candidate spaces and poor-quality reporting. While related to the information retrieval and entity matching literature, existing methods, including LLM-based approaches, are not designed to search effectively when the underlying data are incomplete and inaccurate. Huifeng Su and collaborators developed a domain adapted machine learning pipeline that leverages textual and numerical information, enabling nuanced matching on incomplete and inconsistent data. Their locally deployable, cost-free solution significantly outperforms LLM–based methods in both search quality and runtime, effectively narrowing search spaces and enhancing human matching effectiveness. In doing so, it advances the design and operation of family reunification service.

Huifeng Su is a Ph.D. candidate in operations at the Yale School of Management, co-advised by professors Lesley Meng and Edieal J. Pinker. Leveraging structured and unstructured data, Su’s research focuses on improving and evaluating the efficiency, quality, and equity of essential health and human services (HHS). Su adapts causal inference and machine learning methods to specific operational contexts to generate actionable insights. Prior to his doctoral studies, Su studied industrial and systems engineering and computer sciences at the University of Wisconsin-Madison.