Problem
Case study / v1.1.0 released synthetic pilot prototype
NiyamGuard
An explainable policy-drift and citizen-service sandbox with auditable rule pathways.
Target user
Policy operators and citizens exploring synthetic public-service workflows.
Solution
A review-first sandbox that extracts synthetic policy changes, models rules, identifies drift, and explains verified decisions without performing real government transactions.
Personal contribution
Designed and implemented the rule/evidence workflow, citizen and operator surfaces, safe simulation boundaries, automated regression suite, and portfolio case study.
Architecture
Boundaries before buzzwords.
- FastAPI policy and service APIs
- SQLAlchemy persistence and synthetic fixtures
- React citizen/operator workflows
- Hybrid rule and assistant boundary
Main workflow
State moves with a reason.
- 01
Ingest a synthetic circular
- 02
Extract and review proposed rules
- 03
Compare versions and detect drift
- 04
Publish a verified sandbox rule
- 05
Explain eligibility with evidence
Technical decisions
The trade-offs that define the system.
- Deterministic rule evaluation remains authoritative
- Assistant output is bounded by verified policy evidence
- External identity/payment/government actions stay simulated
Difficult engineering challenge
Keeping citizen guidance helpful while preventing generative output from becoming an unreviewed policy authority required a clear rule/evidence boundary.
Verification
Evidence and boundaries.
Testing
- 243 backend tests
- 60 frontend tests
- Vite production build
Security and privacy
- No real government identity, payment, messaging, or submission is claimed
- Optional geolocation requires user action
Known limitations
- Synthetic policies and service data
- No verified production government integration
Focused future work
- Broaden reviewed fixtures and accessibility/browser evidence
Verified demonstration
Rendered product behavior.
Loading video metadata…