MB.22.31°N · 73.18°E

Software engineer · Security researcher

  1. 01Calibrating orbit loom
  2. 02Weaving glass ribbons
  3. 03Loading case studies & writeups
  4. 04Signal locked

000%

Skip to content
MB.
Back to Projects

// Project case study

Civic Sathi

A multi-role civic platform that connects citizen reports, map context, municipal intelligence, registered contractors, and a protected administration boundary.

Role
Full-stack developer · Team Vertex (SIH 2026)
Duration
Aug – Sep 2026
Stack
ReactTypeScriptLeaflet GISFastAPISQLAlchemyPostgreSQL (Neon)Vercel + Render

The Problem

Civic complaints are triaged by hand, duplicated across channels and rarely tracked to resolution.

The Approach

Four role-segregated portals with vision-based triage, embedding-based duplicate detection, a database-aware copilot, sealed-bid tendering and GIS analytics.

The Outcome

A working multi-portal prototype covering the full lifecycle — report, triage, tender, proof of work and audit.

Product film — Civic SathiPlays when in view

Built by Team Vertex for Smart India Hackathon 2026.

Who it's for

Citizens reporting local problems, municipal officers triaging civic work, registered contractors handling eligible projects, and platform administrators governing the system.

The challenge

Civic services are fragmented across informal messages, maps, departments, and vendors. The product must make reporting easy for residents while giving authorised teams the right context without collapsing every role into one unsafe public dashboard.

What I built

Civic Sathi separates the experience into a public citizen portal, a municipal intelligence portal, a registered contractor portal, and a protected admin surface. Citizens can report a problem with location and evidence, track complaints, explore civic coverage, and choose between Vadodara/VMC and Bengaluru/BBMP contexts. Municipal and contractor access remains behind their own authorised sign-in boundaries.

Why it matters

The value is in the handoff: a resident report becomes structured civic context instead of another isolated message. Location, evidence, severity, issue categories, multilingual access, and privacy-protected aggregation create a clearer path from lived problem to accountable municipal work.

How it works

  1. 01A citizen chooses a city context, reports a problem, adds location and evidence, and tracks the resulting civic record.
  2. 02Nearby reports can be grouped into privacy-protected patterns or hotspots without exposing exact private addresses.
  3. 03Municipal officers enter through the authorised Municipal Intelligence portal to work within their city and department context.
  4. 04Registered contractors use their separate portal when a municipality assigns or manages eligible civic work.

Scope note

The public citizen portal is fully inspectable. Municipality, contractor, and admin pages expose role-specific sign-in boundaries; this case study includes their public entry screens but does not claim private post-login dashboard behaviour or expose credentials.

System Architecture

  1. 01

    Citizens & officers

    4 role portals

  2. 02

    Frontend

    React · TypeScript · Leaflet (Vercel)

  3. 03

    Backend

    FastAPI · SQLAlchemy · ML (Render)

  4. 04

    Database

    PostgreSQL (Neon) · Alembic

  • A public citizen portal with Civic Map, Report Problem, My Complaints, language selection, install-app affordance, and city switching.
  • A visible four-step reporting model: Report, Location, Evidence, and Track.
  • Public issue taxonomy covering water, roads, drainage, sanitation, streetlights, parks, and encroachment.
  • Separate authenticated municipality and contractor entry points, with the super-admin surface intentionally kept private.

// Learnings

What this project taught me

  • 01Deduplication is the AI feature that saves the most human time.
  • 02Strict role segregation is a security feature, not just UX.
  • 03Copilots are only trustworthy when they answer from live data.

// Related work

AI Voice Assistant

K.A.L.K.I.

Desktop assistants either do too little or open an uncontrolled bridge into the host machine.

Lead Intelligence Platform

Neurix Intelligence

Lead lists are stale, unverifiable and expensive — and outreach rules differ by region.

Next project

Neurix Intelligence