AMANA
AMANA is a bounded AI co-worker for emergency departments. It follows every referral across time and institutions, names the next required action and who owns it, and keeps watching until a real-world outcome is evidenced.
Working prototype on synthetic data. Not approved for clinical use.
Start a pilot conversationEvery active referral: where the patient is, how sick they are, what they are waiting for, and how long they have waited.
A specific action, a specific person, a clock — not a notification stream.
The evidence behind every claim, the authority it has to act, and the human decision that granted it.
Pre-authorized internal workflow steps can progress in the synthetic prototype. Clinical decisions and external clinical communication require explicit human authority, and every action is verified against its intended outcome.
AMANA never diagnoses, prescribes, accepts a referral, changes triage, or decides transfer readiness.
NEWS2, risk trajectory and data freshness are computed by code that can be read and tested — never produced by a language model. When inputs are missing, AMANA says so and names them rather than estimating.
AMANA has a read-only hospital-integration foundation with source-contract, conflict and quarantine controls verified using synthetic sources. This is implementation evidence, not a claim of a live hospital connection or clinical deployment.
| Product | Working prototype, synthetic data end to end |
| Engineering evidence | Current review: 542 backend tests, strict typing, lint and frontend checks passed |
| Hospital integration | Read-only ingestion foundation, source-contract validation and conflict checks tested synthetically |
| Clinical status | Not approved for clinical use, not in clinical deployment |
| Next milestone | A governed read-only shadow pilot that observes and measures, and contacts nobody |