ClaimChain · Independent product engineering · 2025–2026
Working prototype
A claims workflow with explicit authority.
ClaimChain tests a product premise: providers, administrators, and buyers should be able to move a claim from intake to purchase without treating interface state, payment redirects, or model output as proof. The working test-data prototype implements the complete path while keeping consequential decisions explicit, versioned, and enforceable.
ML may suggest packages. It cannot approve claims or grant access.
- Product flow
- Provider intake, administrator review and governed packaging, buyer purchase, and entitled export.
- Governing principle
- The backend remains authoritative for eligibility, lifecycle, payment state, ownership, and export access.
- Evidence
- A working three-role prototype with test payments, audit history, versioned rules, advisory ML, and an AWS staging workflow.
See the prototype in action.
A 3½-minute walkthrough of the provider, administrator, and buyer workflows, using test data.
The product question
The core difficulty is preserving eligibility, ownership, payment, and export rights when requests are retried, webhooks arrive later, and an advisory service may return an incomplete or incorrect suggestion.
The product therefore treats authority as part of the user experience: every consequential state should have an explainable source and a bounded path to change.
Where authority lives
- Rules and human review
- Versioned scoring and packaging rules establish eligibility and explain each result; administrators retain the approval and override responsibilities that require judgment.
- Payment and entitlement
- A browser redirect cannot mark inventory sold. Stripe webhook reconciliation establishes payment before the backend grants the buyer access to an export.
- Machine learning
- ML can propose package compositions only after deterministic eligibility checks. It cannot approve claims, bypass constraints, finalize a package, or grant access.
What the prototype proves
The implemented path connects account approval, structured claim intake, document handling, administrative review, explainable scoring, governed packaging, an anonymized buyer view, test purchase, audit activity, and entitled PDF export.
- Explainable under change
- Ruleset versions and recorded decisions make it possible to identify which policy produced a score or package after the rules evolve.
- Safe degradation
- The advisory layer can be unavailable without weakening the authoritative workflow or its deterministic fallback.
The boundary of the work
ClaimChain is a working portfolio prototype, not an operating claims marketplace. It demonstrates the product flow and authority model; production operation would be a separate phase requiring domain validation, legal and compliance review, privacy controls, security hardening, and an operating organization.