My research interests lie at the intersection of human genetics and clinical AI. I aim to harmonize diverse data sources, from electronic health records to multi-omics, to understand the genetic basis of disease and advance patient care.
How can we make AI useful, efficient, and safe enough for everyday patient care?
Cahoon JL, Stanwyck CO, et al.
Building a benchmark to evaluate how well language models retrieve information from electronic health records and reason across a patient's longitudinal history.
Ronaghi S, Tonekaboni S, Stempfle L, Utti V, Cahoon JL, Hendrix N, Vala A, Ghassemi M, Alsentzer E.
A framework for evaluating patient information leakage from medical language models under realistic clinical privacy threats.
Cahoon JL, Stanwyck C, Aali A, Madding R, Sun E, Jiang Y, Dhanasekaran R, Alsentzer E.
TRACE removes templated and duplicated content from clinical notes to reduce language model costs while preserving information for downstream clinical tasks.
Aali A, Mohsin MA, Bikia V, Singhvi A, Gaus R, Bedi S, Cui H, Fuentes M, Unell A, Mai Y, Cahoon JL, Pfeffer M, Daneshjou R, Koyejo S, Alsentzer E, Potts C, Shah NH, Chaudhari AS.
Studying how structured prompting changes benchmark performance and model rankings, using a reproducible framework that integrates DSPy and HELM.
Cahoon JL, Garcia L.
Evaluating wearable-based stress detection across three datasets and comparing gradient boosting with other machine learning methods for continuous monitoring.
How can advances in data collection and processing help us understand genetic disease?
Ungar RA, Cahoon JL, Reuter C, Carter J, Bonner D, Cho M, Montgomery SB, Alsentzer E, Halley M.
Investigating how phenotyping from participants, clinicians, and electronic health records affects rare disease diagnosis.
Fan C, Cahoon JL, Dinh BL, Vecchyo DO, Huber C, Edge MD, Mancuso N, Chiang CWK.
Inferring population demographic history from genealogical trees through a likelihood-based framework.
Cahoon JL, Rui X, Tang E, Simons C, Langie J, Chen M, Lo YC, Chiang CWK.
Characterizing imputation accuracy across ancestrally diverse populations and improving recovery of population-specific variants through meta-imputation.
Additional work in immunology, machine learning, and computing for social good.
Jordan Cahoon. Outcomes Analysis, Stanford University.
Applying causal inference to study the relationship between sanctuary policies and healthcare utilization among agricultural workers.
Jordan Cahoon. Ellison Institute for Transformative Medicine.
Developing models for breast and prostate cancer diagnosis and refining a cloud-based quality control pipeline for whole slide images.
Jordan Cahoon. CSCI 499, University of Southern California.
Evaluating multilingual language models on Spanglish and comparing fine-tuning on code-switched and monolingual data.
Jordan Cahoon. CSCI 461, University of Southern California.
Analyzing predictors of substance abuse referrals for young people aging out of the U.S. foster care system.
Jordan Cahoon, et al. University of Southern California.
Developing a reinforcement learning agent to select and place crops in polycultures for improved sustainability and yield.
Oyesola OO, et al.
Analyzing single-cell RNA sequencing data to characterize how the PGD2–CRTH2 pathway regulates intestinal epithelial responses during helminth infection.
Jordan Cahoon. Baliga Lab, Institute for Systems Biology.
Developing an electroporation protocol for CRISPR-Cas9 delivery and investigating lipid production under nitrogen starvation.