Cancer-center data, governed from source to answer.
Cairn is the layer we would add on top of the Fabric and Power BI estate you already run. It gives you one catalog of certified data products, lineage you can see, research releases with honest-broker controls, and AI that answers only from data you are allowed to see.
Live lineage · the Longitudinal Oncology Patient Record, from the EHRs and registry to research and reporting
Synthetic dataSource · Systems of record, untouched
- EHR (unified, current)HL7 v2 · FHIR R4
- Legacy EHRDecommissioned 2026
- MariaDB EDWExisting warehouse
- Registry softwareNAACCR XML
Bronze · Raw landing, append-only
- Clinical messageslh_bronze · HL7/FHIR
- Legacy bulk extractOne-time, checksummed
- EDW replicaNightly CDC
- Registry dropMonthly
Silver · Cleaned and conformed
- Tumor Registry CasesExtract
- Patient & encounterConformed, MPI-linked
Gold · Certified data products
- Longitudinal Oncology Patient Record · selectedData product
Serve · Reports, research and AI
- Honest-broker datasetsStudy IDs, date-shifted
- Power BI · Oncology qualityClinical leadership
What the concept proves
Start here
The story in four steps
Each step answers one RFI section. A dock at the bottom of each step takes you to the next.
- 1RFI 7.2 · 7.16See the whole estateEvery product, from source to bronze, silver, gold and use, on top of the Fabric and Power BI you already run.Open
- 2RFI 7.3 · 7.6Trust one data productDefinition, lineage, quality checks, owner and an access request that is routed and logged.Open
- 3RFI 7.12Release research data safelyReal HMAC study IDs that differ per study, a k-anonymity check, and re-identification only with an independent approval.Open
- 4RFI 7.14Ask, and be told no when it mattersAnswers only from certified products, with SQL and provenance. A request for patient names is blocked before any SQL exists.Open
Built on what is already there
The platforms named in the RFI stay in place. Osonwanne Group acts as solution architect and integration authority, with open standards (Delta, OMOP CDM, FHIR, SQL) so nothing is locked to us.
- Microsoft Fabric · OneLake
- Power BI
- MariaDB EDW
- SQL Server
- StrataJazz
- Research platforms
- EHR in transition
Real in this demo
- HMAC-SHA256 study IDs, computed in your browser with WebCrypto
- The cohort funnel, date shifting and k-anonymity check, run on a generated cohort
- Policy checks, lineage tracing and quality scoring, all unit-tested
Not real
- Every patient, MRN, product, owner, count and query result
- The language model: questions map to pre-written SQL
- Connections to Fabric, Power BI, Entra ID, an EHR or a key vault