Solvia
A from-scratch platform for laboratory revenue-cycle management — turning a lab's test orders into clean insurance claims and collected revenue across eligibility, prior authorization, coding, claim scrubbing, remittance, denials, and patient billing.
Why I built it
Lab revenue cycle is a long, fragmented pipeline: eligibility (270/271), prior authorization (278), charge capture, coding, claim scrubbing, submission, remittance posting, denials, appeals, AR follow-up, and patient billing — spread across a patchwork of payers, plans, and clearinghouses. The incumbent platforms are entrenched and widely disliked, and labs routinely run their own manual reconciliation to catch revenue that would otherwise slip through. I wanted to see whether the core of that system could be rebuilt to be reliable and transparent, with migration off the incumbent as a real, tested feature rather than a services engagement.
What I built
- A two-tier data model. A Data Vault 2.0 raw vault captures every fact with full lineage and absorbs late-arriving, staggered, non-snapshot extracts without breaking referential integrity. Virtualized star-schema marts sit on top for query ergonomics.
- Multi-tenancy with hard isolation. Schema-per-tenant, with tenant isolation treated as a security-critical property that’s tested on its own.
- EDI / X12 integration across the claim lifecycle, plus a payer-rules engine for the per-payer, per-plan special-casing that revenue cycle lives and dies on.
- Migration as a product. Tooling to import a lab’s existing accounts-receivable extract and stand them up with verifiable parity; round-trip parity tests are the headline of the test strategy.
- Compliance by design. PHI data-flow classification, encryption, least-privilege IAM, audit logging, retention and disposal, and a documented path to SOC 2 / HITRUST.
- Infrastructure as code, end to end. Multi-account AWS org, EKS, Terraform with a “zero unmanaged resources” rule, and CI/CD that gates IaC plan/apply and captures compliance evidence.
Decisions I’d defend
- Data Vault for auditability and messy loads. Real revenue-cycle data arrives out of order and in pieces. A raw vault built for that beats a tidy schema that assumes clean snapshots; the star marts give the ergonomics back on read.
- Migration-parity tests as the headline. If you can’t prove the migrated numbers match the source, nothing else matters. The hardest and most important tests are the round-trip ones.
- An explicit seam for the analytical backend. The architecture names the place where the analytics layer can later move onto Stele (my bitemporal database project), so the audit and time-travel story can extend end to end.
- Compliance as a design input, not a final checklist. PHI flows and tenant isolation shaped the data model and the infrastructure from the first commit.
Honest status
This is an ambitious personal build, not a shipped product. The substance so far is in the architecture, the data and migration core, and the compliance/infra foundation. The genuinely hard parts — referential integrity under staggered extracts, reconciliation, and tenant isolation — are exactly where the work is.