Neurosurgeon · Physician-Scientist · Clinical AI · Global Health

During neurosurgical residency I was already building computer-vision models for intracranial hemorrhage on CT. By late 2022, early language models began extending what I could do in clinical research without a dedicated computational team.

One project grew into a reusable quantitative pipeline that offered a new way to examine normal-pressure hydrocephalus. It remains in use as new patients enter the study in Turkey, and global collaborators independently extended both the dataset and the method into new Hydrocephalus Society work, including an award-winning project.

In 2024, I co-authored a Clinical Neurology and Neurosurgery study evaluating ChatGPT on brain-tumor diagnosis and treatment planning. I built the API-based system that sent clinical questions to the model and captured its answers for independent assessment by practicing neurosurgeons.

With Randy D’Amico, I later moved from benchmarking performance to a harder question: what happens when AI becomes part of the clinical cognitive system? That became our 2026 Neurosurgery letter on the extended mind and the uneven frontier of clinical AI.

It also led to an invited Neurosurgery review on how AI studies should handle data quality, leakage, validation, interpretability, reproducibility and clinical meaning.

With A. Gabriella Wernicke and Netanel Ben-Shalom, I built BrachyAtlas to combine anatomy, imaging, implant geometry and radiation dose in one patient-specific intracranial brachytherapy workflow.

It integrates those inputs with an AI agent that helps the clinician work through voluminous connectomic and expected radiation-damage reports, surfacing the anatomy, likely network effects and treatment tradeoffs that matter most for planning.

During my PhD at the Feinstein Institutes, working with Stavros Zanos and Eric H. Chang, I built a human peripheral-nerve imaging platform with AI as part of the process. Over three years, it grew to span tissue clearing, light-sheet microscopy, molecular anatomy, image analysis and single-fiber reconstruction.

AI also helped overcome physical constraints: I designed and iterated custom holders, access tools and medical-device concepts, then 3D-printed and tested them experimentally. Some concepts became the basis for provisional patent work—bridging ideas, code and physical tools.

The work led to invited presentations at Rockefeller University and the Center for Pain Studies at UT Dallas.

From the same program came the human vagus atlas: we turned thousands of anatomical branch observations into a standardized, surgically relevant map, a Brain Stimulation paper, open data and analysis, and an interactive explorer. NIH SPARC invited a dedicated seminar; Kevin Tracey called it “a major advance in the field of vagus nerve science,” noting the question had persisted “ever since the time of Galen.”

With Randy D’Amico’s brain-tumor group, I build systems that turn dense patient-specific connectomic data into interpretable phenotypes and longitudinal trajectories.

The Disconnectome Explorer is the quickest to try: place a tumor in a 3D brain and it returns the tracts it would disrupt. What began as analysis code has grown into tools the team can use directly—small analytical capabilities that can be called, combined and reused through increasingly agent-integrated workflows.

With Netanel Ben-Shalom, we converted visual cranial-deformity judgment into an automated 3D quantitative step. The work was presented orally at CNS and is accepted in Operative Neurosurgery.

The collaboration led to my involvement with the Society for Neuroplastic Surgery and tools its members can use. I built Neuroplastic Anatomy for publication-quality, reusable AI anatomy illustrations; outputs are entering Society educational supplements. We are also developing transcranioplasty-ultrasound workflows through sonolucent implants.