PhD Student · Master in Law Student
Medical imaging AI, cancer care, and global health
About
I am a PhD student in Bioengineering–Medical Physics at the University of Pennsylvania, advised by James Gee, and a Master in Law student at Penn Carey Law, focusing on healthcare and artificial intelligence.
My research develops and evaluates AI for radiology, radiation oncology, and patient-centered care. I work on medical imaging foundation models, multi-agent systems, and the evaluation methods needed to understand where these tools succeed, where they fail, and how they fit into clinical workflows.
I have conducted research at Penn since 2021 and am an Innovation Fellow at the Penn Center for Cancer Care Innovation (PC3I). As an AI Scientist with RAD-AID International, I contribute to global radiology initiatives, including work with the Botswana–UPenn Partnership. My law studies complement this work through a focus on the governance and institutional decisions that shape healthcare AI.
Research
I study how AI interprets medical information, how its outputs should be evaluated, and what it takes to make these systems useful in practice.
Evaluating spatial and anatomical reasoning in vision-language models, with attention to failure modes across imaging modalities and anatomical regions.
Building modular workflows that separate information extraction, reasoning, generation, and evaluation so that intermediate decisions can be inspected.
Studying radiology report simplification, patient education, and AI applications in oncology and interventional radiology.
Developing reporting practices and studying the technical, legal, and organizational conditions needed for responsible use of healthcare AI.
Benchmark · 2026 preprint
A benchmark and evaluation pipeline for spatial and anatomical reasoning in radiology vision-language models. It includes 300 image–question pairs spanning CT, MRI, and radiography, with analyses across modalities and anatomical regions.
Open-source framework · 2026 preprint
A configurable framework for Multi-Agent Reasoning and Coordination. Role-specific agents pass information through an explicit sequence, making individual stages easier to inspect. MARC supports API-based and local model backends.
Research reporting tool · 2026 preprint
A locally deployable LLM tool that recommends manuscript reporting checklists and helps complete them using evidence from the manuscript. The workflow supports transparent, auditable research reporting.
Global health and funded research
Work with clinical and community partners to address practical needs in imaging access, patient communication, and radiology training.
Google.org Health AI funding · $110,000 team grant
I contribute to a collaboration among Penn Medicine, RAD-AID International, and the Botswana–UPenn Partnership to develop MedGemma-based tools for lung disease care. The planned pilot includes AI-assisted triage, structured reporting, and culturally adapted, multilingual patient summaries, alongside training for local healthcare teams.
The grant was announced by PC3I in April 2026.
Read the project announcementPenn Radiology STIR award · $5,000 · 2025
This project explores an interactive “must-see cases” platform with LLM-based feedback for radiology residents. Supported by a Penn Radiology research grant and mentored by Tessa S. Cook, the work connects clinical education with the evaluation of AI-assisted learning tools.
Publications
Research, methods, and perspectives on medical imaging AI, patient communication, and clinical translation. Preprints are listed separately from journal articles.
An overview of tools and architectures for building agentic systems in clinical AI and radiology.
An evaluation of radiologists’ preferences for the format and length of generated clinical histories.
A systems perspective on AI-enabled dictation and radiology workflow in India.
A perspective on radiologists’ role in shaping the development and adoption of AI.
A perspective on agentic AI and its implications for radiology.
An argument for reporting negative findings to strengthen transparency in LLM research.
A reporting checklist for the development, evaluation, and assessment of LLM studies.
A framework for evaluating the readability and reliability of patient-facing radiology reports.
A perspective on the potential and practical considerations of open-source LLMs.
Governance and implementation considerations for LLMs in health systems.
Manuscripts shared before peer-reviewed journal publication.
A structured prompting method for explicit, interpretable task reasoning.
A synthetic chest CT report dataset for radiology language-model research.
Background
University of Pennsylvania
Bioengineering–Medical Physics
Advisor: James Gee, PhD
Expected 2031
University of Pennsylvania Carey Law School
Focus: Healthcare and AI
Dean’s Scholarship
Expected 2030
Drexel University
Interdisciplinary Studies
Focus: Computer Science, Public Health, and Medicine
June 2026
Research across Radiology and Radiation Oncology, including the Center for Practice Transformation with Tessa S. Cook, global health work with Farouk Dako, and the McBeth Lab with Rafe McBeth. Innovation Fellow at PC3I since June 2025.
Research experience with Christopher P. Bridge and the Quantitative Translational Imaging in Medicine laboratory at the Martinos Center, and with Dania Daye in vascular and interventional radiology. Earlier work includes the Zitnik Lab and Harvard Health Systems Innovation Lab.
R&D internships at Subtle Medical in summer 2025 and summer 2026, studying vision-language models for medical imaging. Research at Drexel’s Sparse Coding Lab with Edward Kim focused on biomedical AI and learning in resource-constrained settings.
Updates
Selected for the Lipman Family Fellowship at the Wharton School.
Pursuing doctoral training in Bioengineering–Medical Physics and a Master in Law at Penn; awarded a Penn Carey Law Dean’s Scholarship.
MARC v1 preprint introduces an open-source framework for multi-agent clinical AI reasoning and coordination.
SPARC-Rad preprint introduces a benchmark for spatial and anatomical reasoning in radiology vision-language models.
Completed my BS in Interdisciplinary Studies at Drexel University.
CheckSupport preprint describes a local LLM workflow for manuscript checklist selection and completion.
PC3I announced our team’s $110,000 Google.org Health AI grant supporting planned work in Botswana.
Speaking and education
Selected lectures, conference sessions, and conversations on imaging informatics, global health, and AI evaluation.
Service and recognition
Contact
I welcome conversations about medical imaging AI, cancer care innovation, global health implementation, and responsible evaluation. For research collaborations, speaking invitations, or questions about the projects here, please email me.
satvik.tripathi@pennmedicine.upenn.edu
Philadelphia, Pennsylvania