Research — the ideas endure.

Code is becoming a commodity; ideas are not. My work sits where machine learning meets a genuinely hard problem — a moon's clouds, a genome's resistance, a failing heart — and asks what mathematical structure makes the problem tractable. The tools change every year. The thinking is the durable part.

Below: three research arcs, a strand of open-source tooling, and the through-line that connects them. Full list on Google Scholar.

Arc I

Planetary science, with NASA

Teaching machines to see across the solar system — instance segmentation and transfer learning applied to spacecraft imagery, so scientists can measure in seconds what once took months. Work with a NASA Goddard planetary-science team.

Rapid automated mapping of clouds on Titan with instance segmentation
Yahn, Trent, Duncan, Seignovert, Santerre, Nixon · JGR: Machine Learning & Computation, 2025
Grid-oriented normalization for analysis of spherical areas from 2-D imageryUS Patent 12,620,126
Nixon, Yahn, Trent, Santerre · granted 2026
Enhanced detection of Martian dust devils from rover images using machine learning
Hatfield, Santerre, Nixon · JGR: Planets, 2026
Detection and segmentation of ice blocks in Europa's chaos regions using deep learning
Europa chaos-terrain work · LPSC / AGU / DPS, 2022–2023

Arc II

Genomics & antimicrobial resistance

The genotype-to-phenotype problem: predicting the behavior of an organism from its DNA. My most-cited work, built during my PhD, brought machine learning to antibiotic-resistance prediction at scale.

Antimicrobial resistance prediction in PATRIC and RAST
Davis, Boisvert, Brettin, … Santerre, et al. · Scientific Reports, 2016 · cited 288×
Machine learning for antimicrobial resistance
Santerre, Davis, Xia, Stevens · arXiv:1607.01224, 2016
Machine learning for the genotype-to-phenotype problem
Santerre · PhD thesis, University of Chicago, 2017 (advisor: Rick Stevens)

Arc III

Applied ML — health, signals & the world

Machine learning pointed at practical, human problems, much of it done alongside graduate students I advise: physiological signals, medical imaging, and a long tail of applied studies.

sEMG gesture recognition with a simple model of attention
Josephs, Drake, Heroy, Santerre · Machine Learning for Health, 2020 · cited 61×
Quantification of ECG instability prior to cardiac arrest in single-ventricle physiology
Savorgnan, Crouthamel, Heroy, Santerre, Acosta · Journal of Electrocardiology, 2022–2023
Deep learning image analysis of S-phase stages in human cells
Boyd, Mitra, Santerre, Sansam · SMU Data Science Review, 2023

Also

Open-source AI tooling

Ideas become useful when they ship. I've contributed to widely used open-source AI libraries — the kind of practical tooling that puts capable models in more developers' hands.

aisuite — a unified interface to many LLM providers
Contributor · one string swaps provider & model across OpenAI, Anthropic, Google, AWS, and more
translation-agent — agentic machine translation
Contributor · a translate → reflect → refine workflow for steerable translation

The through-line

Whether the data is a spacecraft's camera, a bacterial genome, or a child's heartbeat, the move is the same: find the structure the problem is hiding, and let the model exploit it. That belief — that the mathematics is the enduring layer, and code is increasingly the commodity around it — is what I teach, what I build, and what I look for in the problems worth taking on.

PhD, University of Chicago h-index 8 peer-reviewed across 3 fields US patent holder NASA research advisor

Full publication list on Google Scholar →