About · RFI #27-28 Enterprise Data Ecosystem
What this concept is, and is not
Osonwanne Group LLC built this demo to make its response to Roswell Park Comprehensive Cancer Center RFI #27-28 concrete. It shows how the governance, research and AI layers we describe would behave for the people who use them. It is a concept with synthetic data, not a product, not a Roswell Park system and not past performance.
Screens
| Screen | What it shows | RFI |
|---|---|---|
| Estate overview | Medallion layers, product health, freshness against SLA, and what needs attention. | 7.2 · 7.16 |
| Catalog | Search and filter by status, domain, sensitivity and layer. Quality scores are computed from each asset's checks. | 7.6 |
| Data product | Definition, lineage, quality checks, schema contract with PHI columns, owner, steward and a routed access request. | 7.3 · 7.6 |
| Lineage | Source to bronze, silver, gold and use, traced upstream and downstream for any product. | 7.2 · 7.10 |
| Honest broker | IRB request, cohort funnel, HMAC-SHA256 study IDs, date shifting, k-anonymity check, release, logged re-identification. | 7.12 |
| Governed AI | Six policy checks before any SQL, a certified-only allow-list, a blocked PHI request, SQL and a provenance record. | 7.14 |
| Audit log | Every request, decision and query in one searchable, exportable log. | 7.8 |
Real
- HMAC-SHA256 study IDs and date shifts, computed in your browser with WebCrypto using demo keys
- The cohort funnel and k-anonymity check, run on a deterministic synthetic population
- Policy checks, lineage tracing, freshness and quality scoring, each covered by unit tests
- The SHA-256 hash of the SQL in each provenance record
Simulated
- The language model: supported questions map to pre-written SQL and synthetic results
- Fabric, Power BI, Entra ID groups, the EHR, the key vault and the catalog product
- Platform telemetry, approvals and the seeded audit history
How it maps to a real build
The screens show behaviors, not a product choice. In production the catalog, glossary and lineage would run in Microsoft Purview or OpenMetadata (chosen in discovery), data in open Delta tables in OneLake, access through Entra ID and AD groups, and reports in Power BI. Osonwanne Group would act as solution architect and integration authority over those tools, with open standards (Delta, OMOP CDM, FHIR, SQL) so no layer is locked to us. HIPAA Expert Determination is made by an independent qualified expert; the k-anonymity check only supports it.
Contact
Boniface Osonwanne, Chief Executive Officer, Osonwanne Group LLC · bon@osonwanne.com