Three things a web-scraping model can’t give you.
Curation is slow and expensive — which is exactly why it’s the moat.
A hand-built graph
150,000 references reviewed and mapped since 2015 — peer-reviewed sources, not a crawl of the open web.
Graded, not asserted
Each association carries a strength-of-evidence grade, so weak signal is never dressed up as strong.
Traceable to source
Tap any recommendation and follow it to the primary reference (PMID / DOI) behind it.
How an association earns its place.
Every link in the graph clears the same checks before it can shape a recommendation.
- ✓ Drawn from peer-reviewed primary literature
- ✓ Graded for strength of evidence
- ✓ Calibrated to ancestry where the data demands it
- ✓ Curated by a qualified reviewer, not auto-ingested from the web
- ✓ Re-checked as the evidence base changes — associations can be downgraded
A decade of compounding evidence.
The graph is the asset the reasoning runs on.
The honest part.
We don’t invent citations, and we don’t state a regulatory certification we don’t hold. Where evidence is thin, the report says so and lowers its confidence — honest degradation beats a confident wrong answer.
Everything here is decision-support for a qualified professional — never an autonomous diagnosis or a substitute for clinical judgement.
Good reasoning starts with good measurement.
A curated graph is only half the integrity story — the other half is insisting the data it reasons over was measured by a certified lab.
- ✓ Inputs come from certified labs; thin or out-of-range data is flagged and down-weighted, not reasoned over as if it were sound
- ✓ 150,000 curated references, graded for strength of evidence and traceable to a primary source
- ✓ Where evidence or input quality is thin, confidence is lowered and stated — no invented citations, no certifications we do not hold
See the evidence behind a real recommendation.
Open a sample profile and trace any finding to its source.