Sampurna District · Public ledger · FY 2025-26
Where the budget goes should match what people need.
BudgetMirror holds three mirrors up to every ward — citizen needs, budget allocations, and ground outcomes — then shows, in plain numbers, where spending drifts from need.
Anonymous by default · Hindi, Marathi, Kannada, Tamil, Telugu + English · No Aadhaar, no phone
- Citizen reports mirrored
- 390
- Wards covered
- 7
- Critical mismatches
- 2
- Citizen validations
- 6
How the mirror works
Three reflections, one honest picture. AI translates and categorises — but every money score is deterministic arithmetic you can audit.
Citizens speak
Voice or text, in your language. Anonymous by default. AI translates, categorises into 10 civic sectors, and scores urgency — with a graceful fallback if AI is ever down.
Budgets face the mirror
Allocations are uploaded as CSV or PDF, reviewed by officials (never auto-trusted), then compared against demand share. Alignment = 100 − gap × 100. That's the whole trick.
Outcomes get verified
Projects move through a public lifecycle — Planned to Impact Verified — and only citizen validations can close the loop. No validation, no victory lap.
“147 water reports against an 11% budget share is not a data gap — it's a decision waiting to happen. The mirror just makes it impossible to unsee.”
BudgetMirror recommendation engine · Ward 5 · Priority 86
Built to be trusted, not just demoed
Deterministic money math
Demand, alignment, priority and impact scores are pure functions — same inputs, same outputs, every time.
AI never auto-spends
PDF/CSV extraction always lands in a human review grid before touching the database.
Anonymous by design
No Aadhaar, no phone, no identity. Tracking IDs replace personal data.
Open method
Formulas, schema, and seed data ship in the repo. Audit everything.