Overview
The proprietary graph holds standardized healthcare entities (providers, organizations, diseases, drugs, clinical trials, and more) linked by typed relationships. Builds are deterministic pipeline jobs — not ad-hoc AI writes.
Layers
- L1 — Foundation: external knowledge graphs used as read-only anchors.
- L2 — Semantic bridge: UMLS / Biolink resolution authority.
- L3 — Raw material: open healthcare datasets ingested via Spark.
- Proprietary: fused graph in Memgraph used by products.
Identifiers
Nodes use stable kg_id values and, where available, canonical CURIE-style identifiers
(for example NPI, MONDO, RxNorm). Prefer projecting explicit property columns in playground queries
instead of returning whole nodes.
Example
MATCH (n:Provider)
RETURN n.kg_id AS kg_id, n.name AS name, n.npi AS npi
LIMIT 20
Totals and schema catalogs are also exposed via public catalog endpoints under
/api/graph/s3/*. Interactive exploration belongs in the
Playground.