The EquiAccess Indexer
A private, offline health tracker and triage indexer for communities without easy access to a clinic, built first for rural sub-Saharan Africa and South America.
Some people grow up in areas where visiting a clinic means hours or even days of travel. Their stories are why the Indexer exists: to bring the first steps of care closer to home.
Tracks health over time
Monitors health information across multiple visits in a private, encrypted local record.
Pathways, AI-assisted
Follows clinical pathways such as OLDCARTS. A fine-tuned Llama 3.2, trained on real patient and doctor conversations, helps phrase the questions. It never makes the decisions.
Private, local & inclusive
Fully offline. Works regardless of literacy level or dialect, with speech in and speech out.
How a visit flows
-
1
Verification and Vitals
Every visit starts with a user verification, both by date of birth and pin/keyphrase. Then, basic vitals are taken.
-
2
Symptoms
These questions are fixed and never change. They are supposed to gage energy level, overall mental state, and overall physical state.
-
3
Guided Questioning
From information gathered in the previous steps, a clincal pathway is chosen. This continues until enough patient information is collected for next steps.
-
4
Flags
Answers are checked against fixed danger-sign rules. Our AI never decides these, although it can help reveal flags.
-
5
EquiAccess Index
A transparent priority score for clinician review. This is based off of flags and is mostly deterministic.
-
6
Report
A session report is generated and weighed against past session. The visit joins the patient's encrypted on-device history.
deterministic engine AI assists phrasing only
What the AI learned from
One rule shapes every choice: no synthetic data, ever. The model learns to phrase questions only from real clinical conversations between real people, used under licenses that permit it.
MTS-Dialog
Real doctor and patient clinical dialogues.
3,500+ real doctor questionsPriMock57
57 primary-care consultations recorded with real clinicians.
1,900+ real doctor questionsWhat to ask, and which signs are dangerous, is written by people from public-domain MedlinePlus references, never trained into the model. The AI never diagnoses, never decides urgency, and every question it writes must clear a guardrail before a patient sees it.
One visit sees a moment. A month sees a pattern.
EquiAccess remembers every visit. Any single week can look fine. The tool watches across weeks and compiles what one visit can't see. It catches things a clinician might not, not because it knows more medicine, but because it knows the patient. Pick someone and step through their month.
Not a diagnosis. The Index is an information-gathering summary for a clinician; cross-visit flags and weights are drafts pending clinical review. Patients are scripted, but every flag and score is exactly what the engine produced.
