Integrating Machine Learning in Liver Transplantation: Implications for Healthcare Policy and Societal Values in AI-Informed Health Systems
Co-advised by researchers from the Law, Economics, and Data Science Group at ETH Zurich, and Computational Precision Health at the University of California Berkeley and UCSF.
Can machine learning predict who receives liver transplants, detect disparities, and improve care distribution efficiency? In ongoing work for my master's thesis, I explore these questions within the context of liver transplantation (LT) to guide AI use in medicine. I developed a pipeline that combines statistical and machine learning methods, including LLM-based information retrieval, to identify psychosocial factors from clinical notes that influence LT decisions. This research aims to reveal the impact of existing policies and practices, supporting unbiased and equitable transplant decisions to ensure the most suitable patients benefit. By decomposing disparities attributable to social determinants of health and comparing existing and ML-based patient prioritization, we aim to facilitate actionable interventions for more equitable LT listing and completion practices, potentially alleviating future disparities. Additionally, our modeling work will uncover how social values and policies shape healthcare. This has broad implications for improving transparency, equity, and outcomes in healthcare systems, and for informing health policy.


Unlocking Public Preferences: How Natural Language Processing (NLP) and Surveys Can Shape Better Urban Policies
This work was done in collaboration with the Spatial Development and Urban Policy (SPUR) group at ETH Zurich.
Understanding public decision-making is crucial for creating policies that boost democratic trust and participation. This paper combines survey experiments with advanced text analysis to explore how people make decisions on controversial urban housing policies, analyzing responses from thousands of participants across major cities. It reveals that people's concerns extend beyond traditional survey questions, focusing on infrastructure and local issues. The findings highlight relationships and inconsistencies between stated and actual preferences, shedding light on unique decision-making patterns. These insights can guide more effective and responsive urban policies, aligning them more closely with public preferences and enhancing community support. Access the working paper.


Using Mobile Phone Data to Assess Public Transit Expansion and Economic Opportunity in Colombia
This work was done in collaboration with the Spatial Development and Urban Policy (SPUR) group at ETH Zurich.
Understanding the impact of new infrastructure on urban mobility and economic opportunities is essential for equitable urban development. This study focuses on the impact of the new cable car system, opened in 2019, which connects the flat areas of Bogotá and disadvantaged neighborhoods in the southern hills to the city center. We analyze how this new transit option affects the movement patterns and economic opportunities of residents in these neighborhoods, compared to a similar area in the city that still lacks direct public transit connections. The findings will provide insights into how improved transit options can transform local economies and enhance accessibility for underserved communities. Initial code for the data parsing stage of project can be found here.


Modeling Cancer Evolution to Understand Emerging Resistance to Therapy
This work was done in collaboration with the David Liu Lab for Integrative Precision Oncology at Dana Farber Cancer Institute (DFCI) and the Broad Institute of Harvard and MIT.
Cancers evolve like ecosystems. Melanoma, a deadly skin cancer, often develops resistance to treatments, posing significant challenges. In a long-term study of a melanoma patient, we traced how the cancer developed resistance to therapy over the years following treatment. This research revealed specific genetic changes and immune interactions that drive resistance, providing insights into how melanoma adapts to evade treatment. These findings could inform better strategies for overcoming treatment resistance in melanoma. Published paper here.