I’m a Medical Physics PhD student at UW–Madison, researching artificial intelligence (AI) for imaging and cancer care. Outside the lab, I’m drawn to motorsport, music, and the outdoors.
Medical Physics PhD student Advised by Dr. Ran Zhang
Research, software, and the life around them—without needing to read the entire archive.
I came to medical physics through a love of building things and a desire to make them useful. That curiosity has taken me from McMaster to radiation oncology research at UAB, and now to a PhD at UW–Madison.
I’m interested in how AI and vision-language models can help us understand medical images and improve cancer care. Outside research, you’ll find another side of me in motorsport, riding, diving, music, and making things.
My collaboration with Dr. Carlos Cardenas at UAB grew from remote research into a visiting-scholar experience in Birmingham. Along the way, I helped connect students and opportunities across UAB and McMaster.
Mentorship, ambassadorship, and teaching are part of how I hope to make a difference, too.
The story behind the collaborationI work on tools that help clinicians make sense of medical images and plan radiation treatment. Here are three questions I’ve explored.
Radiotherapy teams need consistent names for treatment targets. I explored how locally hosted language models can help standardize them, while keeping clinical meaning in view.
The pipeline evaluated 1,000 clinical names. All outputs passed the TG-263 naming rules, but some changed the intended meaning. The finding: rule compliance still needs clinical review.
How can treatment planning better protect healthy tissue? My first-author study compared four Ethos planning configurations for patients with multiple brain metastases.
Across 45 patients, high-fidelity mode with control rings improved conformity and dose falloff, reduced normal-tissue dose, and lowered plan complexity. This was a planning study, not a clinical-outcomes trial.
An automatically drawn organ boundary needs to be useful to a clinician. I compared three deep-learning approaches using both quantitative measures and blinded physician review.
The study used 122 training and 72 holdout CT images, with three physicians reviewing 30 cases. Both AutoML frameworks outperformed SwinUNETR; physicians preferred nnU-Net over MONAI Auto3DSeg.
Co-authored a 40-patient study integrating prior treatment contours into follow-up MRI; average review time fell from 7.97 to 3.95 minutes. Published online in 2025; journal issue in 2026.
First-author, 45-patient evaluation of four template configurations for multi-metastasis SRS planning.
First-author comparison of SwinUNETR, nnU-Net, and MONAI Auto3DSeg with quantitative and blinded physician review.
Earlier biophysics publication from McMaster work.
ASTRO 68th Annual Meeting · poster
Third author; coauthor, not named presenter · Presenter: Yogesh Kumar, PhD
Conference program record
View presentation recordASTRO 68th Annual Meeting · poster
first author and listed presenter · Presenter: Udbhav Ram, BS
Presented
View presentation recordOpen research and tools I’m building. These are research projects; clinical use requires further validation.
Exploring how specialized AI models make radiology decisions, with open evaluation code and experiments.
A benchmark for testing AI and agent systems on medical-physics tasks.
A macOS tool for turning medical images into 3D-printable research phantoms.
From race data to riding, diving, and making things, I like learning by doing.
In 2023, I took that curiosity to Arrow McLaren as a data-and-strategy intern. My motorsport experience also spans Formula SAE, Formula LGB, and the Polo Cup.
My motorsport journey
PADI Advanced Open Water and Nitrox training built on an earlier SSI Open Water qualification.
Dive into the story
English riding, Western horsemanship, horse care, and time helping at an Alabama lesson barn.
Follow the riding pathWestern and Carnatic violin, piano, and Carnatic singing have been another part of my life alongside science.
Hear the musical threadSelected recognition from research and co-op work, with the full source-linked record kept in one place.
Explore awards, training & recognition
Recognition for the UAB co-op and the broader international research and mentorship path.
View the record →
Designation for one of the highest-scoring AAPM 2025 abstracts, on local language models for radiotherapy naming QA.
View the record →
Undergraduate research competition recognition for the Ethos high-fidelity SRS work.
View the record →The full story, from the newest research to the projects, people, and experiences that came before.
The path from McMaster and UAB to Wisconsin.
Every institution, role, project, and source in one place.
The complete record, with ownership and evidence labels.
Papers, posters, presentations, and source material.
Open research tools, engineering, and experiments.
Profiles, conversations, and videos from work and life.
Motorsport, music, diving, and the other parts of life.
124 records, organized and searchable.
