Skip to content
JANSETU AI
FROM VOICE TO ACTION
DIGITAL PUBLIC GOOD

Infrastructure is not about concrete. It is about dignity.

JANSETU AI is not a complaint chatbot. It is a public infrastructure demand intelligence layer connecting citizen demand to evidence, prioritization, investment planning and measurable outcomes.

Citizen-firstPrivacy-preservingFairness-constrained

OUR MISSION

As a Digital Public Good, JANSETU is citizen-first, privacy-preserving, and fairness-constrained — we never use religion, caste, or political affiliation in any scoring or prioritization decision, and every recommendation is traceable to auditable evidence at a pinned weight version.

The platform was built India-first for Gujarati, Hindi, and English citizens who can speak, type, or share a photo, and designed BRICS-ready for Brazil, Russia, India, China, and South Africa with country-specific languages, admin hierarchies, currencies, and datasets. Our pipeline is: Citizen Voice → AI Understanding (Gemini translates and structures category, urgency, location, affected groups) → Evidence Fusion (BigQuery GIS joins demographics, infrastructure indices, investment gaps) → Deterministic Priority Engine v1 (demand 30% + infrastructure gap 20% + population impact 15% + vulnerability 15% + urgency 10% + feasibility 10%) → Candidate Project → Policymaker Evidence Review → Human Decision → Impact Tracking (baseline → target → actual, observed vs modeled). Frontend is untrusted: all validation, clustering, scoring, recommendation, and audit live server-side.

GU · HI · EN
Voice, text, photo
BRICS-ready
5 countries
Deterministic v1
Pinned weights
GOVERNANCE

Human decides. System advises.

Frontend is untrusted. Backend owns validation, scoring, audit. Gemini may understand, classify, translate, explain, draft — must not invent evidence, alter weights, approve funding, or override authoritative data. Every priority score persists every component plus weightVersion for reproducibility.

Human review required
EVIDENCE

Traceable to source

  • Every recommendation cites evidence refs + weightVersion
  • Public analytics use centroids & GeoJSON — never individual positions
  • Audit logs record every approve/reject with reason

PRINCIPLES

Fairness-constrained
No religion, caste, politics in scoring — enforced in code and review.
Privacy-preserving
Voice → text only. No biometrics, no sales. Location opt-in; text locality works.
Deterministic
Same inputs, same score — always. Versioned weights, explainable drivers.
Auditable
Full trace: citizen text → intake → cluster → score → project → decision → impact.
CONTACT & PILOTS

Questions about the platform, partnerships, or BRICS deployment:

EMAIL
support@jansetu.ai
Fastest — 1 business day
PHONE
+91 265 123 4567
Mon–Fri 10–18 IST · GU/HI/EN

Vadodara Innovation Corridor, Vadodara, Gujarat 390001, India · /contact · /privacy · /docs

Contact us Developer docs
Built for hackathon demo on synthetic datasets clearly labeled as such — swap services/api/src/services/store.ts with Firebase Admin + BigQuery GIS for production.