CleanAir & Clear Streets

VayuLens

Neighbourhood pollution intelligence that detects hidden hotspots, predicts AQI spikes, dispatches municipal teams, and proves the air improved.

The Problem

City AQI misses the street where people are actually breathing smoke.

Dump fires, construction dust, industrial plumes, and smog-trap junctions can harm one ward while the official AQI station looks normal.

Why Existing Apps Fail

Most tools stop at a red dot.

Sparse stationsRegulatory monitors cannot see every street or dump edge.
Noisy complaintsCitizen reports need trust, tamper checks, and corroboration.
Coarse satellite dataSatellite products are useful priors, not lane-level truth.
No proof loopAlerts rarely show whether a municipal action worked.
The Solution

An evidence graph for hyperlocal pollution response.

VayuLens fuses citizen photos, low-cost PM sensors, NASA FIRMS fire priors, Sentinel/MODIS context, and weather into Watch, Suspected, Confirmed, and Mitigated hotspot states.

AI Execution

The AI does operational work, not decoration.

Photo intelligenceSmoke, dust, garbage-fire probability, haze proxy, EXIF confidence.
Sensor anomalyHumidity-corrected PM2.5/PM10 with local baseline deviation.
Wind-aware fusionReports become stronger when downwind sensors spike.
24-hour forecastXGBoost now, spatiotemporal GNN later.
Demo Moment
96%

Confirmed dump-fire hotspot

One citizen photo becomes a confirmed incident only after sensor, fire-prior, and wind-consistency evidence agree.

Municipal Action

From detection to dispatch.

The action layer assigns MC-03 mist cannon and SW-12 cleanup crew based on cause, ETA, exposed population, school proximity, and predicted AQI spike.

Accountability

Proof of Action is the headline feature.

Before/after PM2.5, PM10, visual haze, SLA closure, and citizen complaint trends prove whether the intervention worked.

Deployability

Prototype today, Google Cloud city pilot next.

NowStatic prototype on GitHub Pages via GitHub Actions CI/CD.
PilotCloud Run, Pub/Sub, Cloud SQL/PostGIS, Cloud Storage, Vertex AI, BigQuery.
IntegrationCPCB/SAMEER context, Swachhata-style complaints, ICCC dashboards.
ScaleWard-by-ward rollout using CSR-funded sensors and municipal fleets.
Inclusion

Built for Indian city conditions.

Cheap Android support, offline queue, compressed image upload, Hindi/regional labels, voice notes, privacy blur, and low-literacy-friendly workflows.

The Ask

Pilot one ward in 30 days.

20 low-cost sensors, sanitation fleet participation, ICCC dashboard access, and school/RWA citizen reporting.