Data leaves the practice by default
Most consumer AI tools send text and files to third parties. For a physician or a practice, that is a privacy, compliance, and trust problem.
Physician-led AI for individuals and independent practices
Build and operate AI safely. Local models first.
I help individual professionals and independent medical practices build and operate AI on their own terms, with a focus on local models, safety, and human review. No PHI leaving the practice unless you choose it to.
Built by a practicing radiation oncologist. Services, physician education, in-person best-practice lectures, and a weekly newsletter for physicians on AI in medicine.
- Local models first
- PHI-conscious by design
- Human in the loop
- Physician-led
The problem
Most AI advice for medicine ignores where the risk actually is.
Physicians and small practices are being told to use AI without a clear picture of what happens to their data, where the model runs, and who is responsible when it gets something wrong.
The right answer is not to avoid AI. It is to design for safety first, use local models where they fit, and keep humans in the loop for anything that matters.
Nobody explains the safe path
Best-practice guidance is scattered across vendor marketing, academic papers, and blog posts. Nothing speaks plainly to a working clinician.
Small practices get skipped
Enterprise consultants build for large systems. Independent practices need a right-sized approach that fits their staff, budget, and workflow.
How it works
Safety first. Local models where they fit. Humans in the loop.
Every engagement follows the same spine, sized to the individual or practice. We do not push enterprise-scale tooling on a five-person clinic, and we do not oversell what AI can safely do.
Design
Map what data is in scope, what must never leave the practice, and where human review is required. Choose local models by default.
Build
Stand up local models and workflows on your own infrastructure. Use cloud services only where they are safe and clearly needed. Everything stays client-owned and portable.
Operate
Train you and your team to run it. Add review checkpoints, recurring updates, and an audit trail. Nothing here is clinical decision support.
Services
Build and operate AI safely, at your scale.
Services are scoped case by case. Individuals typically start small. Practices start with a scoped engagement and can continue on retainer.
Individual build-and-operate
For high-complexity professionals who want to use AI safely and effectively in daily work. Local models first, private by default, right-sized to your setup.
Independent practice build-and-operate
For small and independent medical practices adding AI to administrative or operational workflows. PHI-conscious, local-first, and paced to what your staff can actually adopt.
Ongoing safety and operations
Optional retainer to maintain your setup, review new tools before you adopt them, and keep humans in the loop as things change.
Education
Physician AI education, from the basics up.
Materials for physicians who want to understand AI on their own terms. Starts with the basics that most vendors skip, then moves into the practical questions that actually matter in clinic.
AI basics for physicians
What a model is, what it is not, what local versus cloud actually means, and how to read AI claims critically. Plain English, no vendor talking points.
Working topics for clinicians
Prompting, workflow design, safety and consent, PHI-conscious use, evaluating vendor tools, and when local models are the right answer.
In development
First modules are being built now. Subscribe to Practical AI for Physicians to be told when they are available.
In-person lectures
Best-practice AI lectures for clinics and facilities.
Bring a working physician on-site to teach your team what safe, sensible AI use actually looks like. Talks are tailored to your setting and audience.
Safety and privacy essentials
What is safe to send to AI, what is not, and how to build simple rules your staff can actually follow. Covers PHI-conscious use in plain language.
Local models for medicine
Where local models fit, how they change the risk profile, and what an implementation for a real clinic looks like end to end.
CME-accreditable, in development
Talks are being prepared for CME accreditation through a partner. Available on request in the meantime as non-CME best-practice lectures. Subscribe to the newsletter to be notified when booking opens.
Safety and privacy
Local models first. Humans always in the loop.
Healthcare AI without guardrails is a liability story waiting to happen. Every engagement is built around explicit boundaries.
- Local models by default. Cloud only where it is safe and clearly needed.
- PHI-conscious by design. No PHI to public models.
- Human review stays required for consequential work.
- Not clinical decision support. It is an operational and administrative layer.
- Client-owned systems. Portable and reviewable.
- No autonomous, irreversible actions.
- Nothing here is legal, financial, or medical advice.
Who this is for
And who it is not.
Built for
- Physicians who want to use AI safely in real work.
- Independent practices adding AI to operational or administrative workflows.
- Clinicians who want plain-English education, not vendor slide decks.
- Facilities looking for a physician-led speaker on safe AI use.
Not for
- People looking for autonomous AI to replace clinical judgment.
- Groups unwilling to keep human review in the loop.
- Organizations expecting HIPAA-graded work over public models.
- Anyone hoping for a shortcut around safety and consent.
Founder memo
Read the founder memo.
The deeper essay explains why local models matter for medicine, why safety and human review are the real product, and how ParallelOS actually shows up for individuals and independent practices.