A day in the field
Follow one record through the whole system.
Press play
Seven moments, from a form answered on a phone at dawn to an AI draft approved in a person's name at dusk. The camera follows the record.
Public health runs on research. Research runs on ten tools.
One platform where the field, the evidence, the grant and the paper are the same record.
AMDana is the operating system for public-health research teams and the field providers who feed them: tasks and projects, messaging, forms and live documents, statistics with provenance, grants with a real submit gate, systematic reviews blinded by construction, IRB compliance — and AI that proposes while a person decides.
The problem
Every seam between tools is where a study loses time, data and its audit trail.
In the field
Data is collected on paper or in a form tool that knows nothing about the study it serves. It is re-typed into a spreadsheet, emailed to a statistician and analysed on whichever laptop was open.
In the office
The grant lives in one system, the IRB approval in another, the tasks in a third and the conversation in a chat app. Nobody can say, from a record, why a number in the manuscript is what it is.
At the reviewer's desk
A funder, a journal or a monitor asks where the data came from, who approved what and when. The answer is a memory and a search of somebody's inbox.
The scale of it
The platform
Eleven districts. One data model. Click any district to fly to it.
The hub is one customer's federation. Every district is a real area of the product; the packets are records moving between them; the phones at the edge are field teams who hold no account at all.
A stylised simulation of the live product, drawn from its own feature list — not screenshots.
AI with a human gate
AI proposes. A person decides. Everything is recorded.
Every AI result in AMDana is a record, never a mutation. A run lands with its provider, model, prompt template and version, prompt and response hashes, structured output, supporting evidence, confidence, token counts and cost. It arrives pending review. Only an approving human applies it — in their own name, with the run linked as the rationale. Rejection changes nothing.
The copilot builds its context only from the caller's visible projects, so it cannot answer with another institution's data. Without credentials the whole layer runs in a labelled sandbox computed from the real records, so every screen works before a single token is bought — and every token that is bought runs through one meter, Smart Credits.
The concierge on this page is the same idea in miniature: it answers from the fact sheet you are reading, and says so when it cannot.
res_ai_runspending_review
- Task
- progress_report_narrative · v3
- Provider · model
- anthropic · recorded per run
- Prompt hash
- sha256:9f1c…e2a7
- Response hash
- sha256:41b0…7c9d
- Evidence
- 4 records the caller can see
- Confidence
- 0.82
- Tokens · cost
- 2,318 in · 611 out · metered
- Applied by
- — waits for a person
Security & trust
Built the way a compliance officer would have built it.
None of this is a roadmap item. Each line describes the software as it runs today.
Business model
Subscriptions the product enforces, AI the product meters, and instances an institution can own.
Why AMDana wins
The moat is the data model and the provenance it carries.
The $5M plan
Five million dollars, twenty-four months, five lines.
The product exists and is in production. This round buys the three things a working platform needs to become the standard for its field: the offline-first field app, the certifications institutions require before they sign, and the go-to-market to reach the schools, centers, ministries and networks that need it.
Proposed allocation. Definitive terms in the data room.
What we would commit to
Targets, not promises. What is not built yet — the offline app, institutional literature connectors, formal certification — is named here on purpose.
Next steps
See it run. Ask it anything. Then let's talk.
The whole product is a 30-day trial away, with no card. The data room holds the terms, the metrics we do not publish here and the security documentation.