AORTA-7B · LM Studio · running locally
AORTA is an open-source framework for turning any capable AI model into a domain-expert colleague for organ procurement. Built by OPO professionals, for OPO professionals. Not a chatbot — a behavioral specification, policy corpus, and training methodology that produces a calibrated, safety-constrained AI partner for the hardest job in healthcare.
Organ procurement coordinators manage life-or-death decisions across 21 chapters of OPTN policy, CMS Conditions for Coverage, institutional SOPs, and hospital protocols — often simultaneously, often at 3 AM. The complexity is increasing. The workforce is stretched. The margin for error is measured in human lives.
Current tools are transactional: data entry, match run interfaces, case management forms. None of them think. None of them know the policy. None of them can reason about the intersection of allocation rules, regulatory requirements, and institutional procedures under time pressure.
AORTA is not a product. It's a complete open-source framework — a behavioral specification, a RAG-optimized policy corpus, a reasoning trace methodology, and a training pipeline — that any OPO can use to deploy AI-augmented coordination support. The framework is substrate-portable: it works whether prompted into a frontier model or fine-tuned into a local deployment.
Every response carries an explicit confidence signal — HIGH when grounded in retrieved policy text, MODERATE when reasoning from domain knowledge, LOW at the knowledge edge. Coordinators always know how much to trust the answer.
AORTA will never make a clinical decision, determine organ viability, override allocation, or replace the judgment of a physician, coordinator, or medical director. This boundary is architectural, not advisory — it's encoded in the model's training, not just its instructions.
Answers cite specific OPTN policy sections. The RAG corpus is structured with semantic metadata for precise retrieval. When AORTA doesn't have the policy text, it says so rather than generating plausible-sounding policy language.
In organ procurement, telling someone what they want to hear can cost a life. AORTA is trained to maintain analytical integrity under pressure — from frustrated surgeons, stressed coordinators, or ambiguous situations where the easy answer isn't the right one.
Each component stands alone. Use the soul document with a frontier model API and nothing else. Or go deep — fine-tune a local model with the full training pipeline. The framework scales to your technical capacity and operational needs.
A policy reference that reasons, not just retrieves. Ask it about the intersection of DCD protocols with allocation rules at 3 AM and get a cited, calibrated answer.
A complete deployment guide from model selection through RAG configuration to system prompt architecture. Works with local inference or hosted APIs.
A framework for evaluating AI deployment in organ procurement — with safety constraints, compliance considerations, and measurable outcomes built in.
A case study in domain-specific AI deployment for safety-critical healthcare operations. The methodology generalizes beyond organ procurement.
AORTA was developed at a US organ procurement organization by a systems administrator who understood both the operational reality of procurement coordination and the current capabilities of open-weight AI models. It emerged from a simple observation: the people who do this work deserve tools that think, not just tools that store data.
This is not a startup. There is no funding round, no sales team, no enterprise pricing page. AORTA is released under the MIT license because organ procurement is a public trust. Every OPO in the country operates under the same OPTN policies, serves the same mission, and faces the same complexity. The tools that help coordinators navigate that complexity should be shared, not sold.