I build models that learn from complex data and the systems that make them useful.
I'm a computer science master's student and researcher in Dartmouth's Hassanpour Lab. My work spans foundation model adaptation, computational pathology, medical computer vision, and robust ML evaluation. I like research that survives contact with the real world.
From gigapixel pathology slides to production-minded AI tooling. A few projects at the intersection of research and engineering.
Adapting Prov-GigaPath representations for liver disease classification, including attention-based multiple instance learning, multimodal fusion, and LoRA fine-tuning experiments.
Most agents are a text box. Bitty is an agent you can see working: the creature reacts to what the agent is actually doing (thinking, running a tool, asking permission, failing) instead of just showing a chat log.
Developed deep-learning approaches to reconstruct stopping-power ratio maps from dual-energy CT, evaluated against Gammex-provided reference maps using SPR percentage deviation.
Investigating whether shortcut-mitigation methods retain their effectiveness across dermatology datasets and skin-tone subgroups, rather than only on their original benchmarks.
Investigating reliability of medical foundation models as data changes across time, imaging modalities, and rare-disease cohorts. My thesis studies temporal drift, synthetic-data feedback loops, subgroup failure modes, and mitigation through calibration and lightweight adaptation.
A terminal-based coding agent exploring retrieval-augmented context, tool use, and file-permission guardrails.
Computational pathology, medical foundation models, and computer vision for clinical imaging.
Interdisciplinary imageomics research and field-based applications of computer vision.
Machine learning for dual-energy CT and proton therapy stopping-power estimation.
Embedded Linux and VxWorks systems, developer tools, networking, and software automation.
I care about the gap between a promising model and a system people can actually trust.
My favorite problems combine messy data, careful experiments, and engineering constraints. I enjoy moving between reading papers, training models, and figuring out why a pipeline broke at 2 a.m.
Outside of research, I ride horses, enjoy a good story, and tend to collect more side-project ideas than my calendar can reasonably support.
currently: learning, building, debugging, repeating ↗
Open to conversations about ML research, AI engineering, and interesting collaborations.