Selected work / PRAMAAN Assurance
Evidence governance · Provenance · Assurance Implementation baseline
Evidence with a history
Designing evidence controls before activating an institutional assurance platform.
Every record. A traceable story.
An evidence archive becomes useful when every record has context, provenance and a clear approval history.
The problem
A metric number or filename cannot establish which framework, reporting period or approved use an institutional record belongs to. PRAMAAN Assurance addresses that gap through a separate application workspace, a read-only inventory pipeline and a governance model for reviewing evidence before migration. The architecture distinguishes source documents from controlled metadata and treats approval as an accountable state change.
My contribution
Product requirements, architecture and documented implementation, produced through an AI-assisted development workflow directed by iterative review.
Deliverables
- Read-only archive inventory and extraction pipeline
- Checksums and data-quality reports
- Relational schema and migration
- Controlled workflow services
- Operating, recovery and test documentation
Process
Decisions that shaped it
Preserve the source.
Keep the original archive separate from the application. Retain complete provenance paths so a reviewer can trace the origin of a record.
Separate evidence from metadata.
Model document objects, structured metadata and audit history as distinct concerns. This makes permissions and changes easier to reason about.
Put authority in the service.
Design authorisation and controlled transitions on the server. A role selected in a browser must not become an authority source.
Outcome
The project provides a documented technical baseline. Its reports record a passing production build, scoped lint and isolated schema-acceptance checks. These support further validation of the design; they do not establish that the service is in institutional use.
Takeaway
Systems thinking that connects evidence integrity, access control, accountable review and honest release decisions.
Honest limits
Hosted identity and storage integration, authorised migration and production validation remain pending. The AI-provider boundary is intentionally unconfigured. This is not presented as an operational AI platform.
What would come next
- Validate hosted identity and storage integration
- Confirm framework and evidence permissions
- Complete operational and recovery testing