Research Group
DermUnbound
Clinical AI, Physician-Controlled
Open-source, privacy-first clinical AI. Built by a physician, for physicians.
Led by Dr. Yehonatan Kaplan — ACMS Fellow, Mohs Surgeon, building the tools he wished existed.
“The tools a physician uses should serve the physician — not the other way around.”
The Problem
Clinical AI is broken
Today's dermatology AI tools live on someone else's servers. Patient images, clinical notes, and diagnostic data flow through cloud platforms controlled by technology companies — not by physicians. The doctor becomes a user, not an owner.
These tools are designed for scale, not for the solo practitioner. They require internet connectivity, ongoing subscriptions, and trust in third-party data handling. For the physician who values patient privacy and clinical autonomy, the current landscape offers no real alternative.
Imagine a Mohs surgeon who needs to process a pathology report during surgery. Today, that means uploading patient data to a cloud service — sending protected health information across networks you cannot audit. With DermUnbound, it means running a local model on your clinic's computer. The data never leaves the room.
We believe there is a better way. Tools that run on your hardware, under your control, with your patients' data never leaving your clinic.
$9.77M
Average cost of a healthcare data breach (IBM, 2024)
87%
Of physicians concerned about patient data in cloud AI tools
Zero
Lines of patient data sent to the cloud by DermUnbound
Our Approach
Three pillars of physician-controlled AI
Privacy-First
All tools run locally or on-premise. Zero cloud dependency. Full air-gap capability. Your patient data never leaves your machine.
Physician-Built
Created by a practicing dermatologist and Mohs surgeon. Every tool solves a real clinical problem encountered in daily practice.
Open Source
MIT and CC BY-NC-SA licensed. Fork it, modify it, deploy it. Docker containers for one-command setup. Full transparency.
Flagship Project
The Docker Framework
A curated Docker-based framework for downloading, running, and managing open-source AI tools tailored for dermatology. One command to deploy an entire clinical AI suite.
# Pull the DermUnbound AI toolkit$ docker pull dermunbound/derm-ai-toolkit# Start all services$ docker compose up✓ Dermoscopy AI model loaded✓ Clinical documentation engine ready✓ Pathology correlation service started✓ Dashboard available at localhost:8080
Dermoscopy AI
Open-source skin lesion classification models
Surgical Planning
Defect analysis and reconstruction options
Privacy Architecture
Air-gapped operation, zero cloud dependency
Local Inference
Run on your hardware, GPU-accelerated
Portfolio
Projects
MohsPedia
LiveClinical decision support for Mohs micrographic surgery — 7 interactive tools
OptiMohs
LiveStep-by-step Mohs surgery wizard with integrated decision support
Claude Academy
LivePractical AI guide for physicians — 7 clinical domains, ready-to-use prompts & protocols
PathCorrelate
LiveAutomated pathology report processing and correlation
5
Production Tools
269
Clinical Templates
4
Research Papers
0
Patient Data in the Cloud
Research
The Physician-Controlled AI Trilogy
From Cloud to Clinic: A Framework for Physician-Deployed AI in Dermatology
In PreparationKaplan Y. · Preprint · 2026
Enterprise Cloud AI in Dermatology: Capabilities, Limitations, and the Case for Local Deployment
In PreparationKaplan Y. · JMIR Formative Research · 2026
Physician-Controlled AI: Privacy-First Deployment Using Docker Containers in Dermatology
In PreparationKaplan Y. · JMIR Medical Informatics · 2026
Vibe Coding in Medicine: How Clinicians Can Build Their Own AI Tools
In PreparationKaplan Y. · JAAD Open · 2026
“Every clinical tool in this portfolio was born from a real problem encountered during a real patient encounter.”
The clinician-coder is not a software engineer who happens to know medicine. It is a physician who uses modern AI-assisted development tools to build exactly what their practice needs — nothing more, nothing less. Every project in this portfolio was built by a practicing dermatologist, during clinical work, to solve real problems.
This is what we call vibe coding in medicine — a term coined by Andrej Karpathy — using AI assistants not as a replacement for clinical judgment, but as a force multiplier for turning clinical insight into working software.