AI-AMP // AI in Austere Medicine Project

AI
for medicine
in the margins

Open source. Built in the field. Safety shared.

We build open source AI tools for medicine where infrastructure ends — and we publish everything, including what fails. Clinical AI gets safer when it's built in the open.

View Projects GitHub Org ↗ Beta Test First Time Guide Research
3×
Open Source Projects
15k+
Lines of Code Written
7k
Knowledge Base Chunks
4
Research Docs Published
Why We Exist

The Gap
Nobody Filled

AI-AMP believes that preventable mortality is highest exactly where clinical AI is least available. Combat medicine, wilderness rescue, austere field care, low-resource hospitals — the environments with the most to gain from decision support are the ones it was never designed for.

Existing tools assume hospital infrastructure. We design for the absence of it. No reliable power. No internet. No physician backup. A single provider who is the entire system.

We are an open-source research initiative building the tools, methodology, and evidence base for AI in austere medicine. Built by clinicians. Evaluated by working providers. Published openly so anyone can build on it.

We document what fails, not just what works — because in clinical AI, understanding failure modes matters more than reporting accuracy scores.

Open by Default
Every line of code, every evaluation report, every safety finding is public. Closed systems hide their failure modes. Open systems let the community find and fix them.
Field Validated
Not a demo. Not a simulation. Real deployments, real providers, real feedback. Field use reveals failure modes that automated testing never will.
Medic-Built
Designed by advanced practice paramedics and clinicians who work in the environments this system is built for.
Affordable at the Margins
Free software, minimal hardware. Every tool targets the lowest possible operational cost, because AI for medicine in the margins has to be affordable in the margins.
What We're Building

Three Projects.
One Mission.

Each project pushes the technical limits of what's possible on off-the-shelf hardware in resource-denied environments. Together they form a living roadmap — open source evidence that others can build on, adapt, and take further. We don't just build tools. We establish what's possible and show exactly how we got there.

PROJECT 01
Active — v3.0
EdgeCDSS
Cloud
v3.0.0 // Two-Pass Safety Architecture
Hybrid cloud-edge clinical decision support. Voice-accessible, protocol-driven guidance in under 3 seconds. All medication dosing resolved to final mL volumes. Two-pass safety validator blocks dangerous responses before they reach the field.
FastAPI GPT-4o-mini ChromaDB GCP ElevenLabs JTS CPGs
View on GitHub →
PROJECT 02
In Development
EdgeCDSS
Offline
Summer 2026 // Zero Infrastructure
Fully offline clinical AI on a $250 edge device. No internet. No cloud. No dependency. Local LLM inference on Jetson Orin Nano. Designed for truly denied environments — conflict zones, remote expeditions, communications blackouts.
Jetson Orin Phi-3 / Mistral Hugging Face Iridium Solar
Follow Progress →
PROJECT 03
Planned
EdgeCDSS
Hybrid
Coming // Technical Frontier
A hybrid of Projects 01 and 02 — pushing the absolute limits of off-the-shelf hardware for denied-comms medical AI. Expanded Iridium satellite integration, LoRa mesh protocols for multi-device field networks, and fully autonomous local inference with opportunistic cloud sync. Built to operate where nothing else does.
Iridium SBD LoRa Mesh Edge LLM Cloud Sync Off-Grid
Follow for Updates →
Open Safety Research
We Publish
Our Failures.

In field evaluation, external clinical testers identified patient safety failures that our automated test suite completely missed. We published all of them. The dosing errors, the missing surgical airway, the dangerous oxygen recommendation — documented, analyzed, and fixed in the open.

This is how clinical AI gets safer. Not by hiding failure modes behind NDAs and closed evaluation processes. By building in public, testing with real providers, and publishing exactly what went wrong and exactly how it was fixed.

Read the evaluation reports →

Weekly Reading

Research
That Drives Us.

Studies, benchmarks, and frameworks shaping AI in austere medicine — updated weekly. Each entry notes how it connects to what we're building. External links, no paywalls where possible.

Loading research entries…
Join the Project

Four Ways
to Help.

🎙️
Beta Test
You're a medic, paramedic, nurse, physician, or rescuer. You understand what the system needs to get right. Test it, break it, flag what's wrong. Your feedback directly drives the next version.
Sign up for beta access →
⌨️
Contribute Code
Python, FastAPI, LLM prompt engineering, ChromaDB, edge hardware. Open issues on GitHub, submit PRs, build something that saves lives in the field.
View open issues on GitHub →
📡
Build Hardware
Off-grid communications, solar charging, edge computing, Iridium satellite, LoRa mesh, WireGuard networking. The offline and hybrid builds need people who know how to make hardware work where infrastructure doesn't.
See hardware roadmap →
🔬
Research
Read and cite the published evaluation reports. Contribute clinical expertise to protocol review. Collaborate on field research methodology. All data is open and reproducible.
Read the research →