Mission
Why we build
A manifesto for physician-controlled clinical AI — why the current model is broken and what we are doing about it.
The Problem
AI adoption in dermatology is failing the solo practitioner
The promise of AI in dermatology has been captured by technology companies building cloud platforms that require physicians to upload patient data to servers they do not control. The business model is clear: aggregate medical data, train proprietary models, and sell subscriptions back to the doctors who generated the data in the first place.
For the solo practitioner or small practice, this model creates dependency without autonomy. You cannot modify the tool. You cannot verify what happens to your data. You cannot operate without an internet connection. You are a consumer, not an owner.
This is not a privacy concern alone — it is a professional sovereignty concern. A physician who cannot control their own tools cannot fully control the care they deliver.
The Vision
AI tools that serve the physician
We envision a world where every dermatologist can download, run, and customize AI tools on their own hardware — without relying on a cloud platform, without signing a data-sharing agreement, and without needing a software engineering degree.
The tools should be as accessible as installing an app. They should work offline. They should be transparent — open source, so any physician can inspect, audit, and modify them. And they should solve real clinical problems, not theoretical ones.
Our Principles
What we believe
Privacy is non-negotiable
Zero cloud dependency. Air-gap capable. Patient data never leaves the physician's machine. This is not a feature — it is a foundation.
Open source is a philosophy, not just a license
Transparency means more than publishing code. It means documenting decisions, welcoming contributions, and building in public. If a tool cannot be inspected, it should not be trusted with patient data.
The physician controls the tool, not the vendor
Every DermUnbound tool can be forked, modified, and deployed independently. There is no vendor lock-in, no required account, no subscription gate. The physician is the administrator.
If a physician cannot install it in 10 minutes, it is not ready
End-user experience matters. Docker containers enable one-command deployment. If a tool requires manual configuration, dependency management, or command-line expertise, it needs more work.
Every tool must solve a real clinical problem
DermUnbound tools are not academic exercises. Each one was created to address a specific clinical need encountered during daily practice in dermatology and Mohs surgery.
The Clinician-Coder
Why physicians should learn to build
The most effective clinical AI tools are not built by technology companies trying to understand medicine. They are built by physicians who understand technology just enough to turn clinical insight into working software.
This does not require a computer science degree. Modern AI-assisted development tools — what we call vibe coding— allow clinicians to describe what they need in natural language and iterate toward working solutions. The physician contributes domain expertise; the AI contributes code generation. The result is software that solves real problems because it was designed by someone who encounters those problems daily.
“A physician who understands their digital tools can deliver better, faster, and more personalized care to their patients.”
Looking Forward
Research agenda
DermUnbound is actively researching three interconnected areas: the clinical utility of locally-deployed AI in dermatology practice, the security and privacy implications of cloud versus local AI deployment, and the feasibility of physician-led software development using AI-assisted coding tools.
Our findings are documented in the Physician-Controlled AI Trilogy — a series of papers exploring the enterprise cloud model, the Docker-based local deployment alternative, and the emerging clinician-coder paradigm.