NagarFlow is an AI-powered platform that predicts demand and optimizes allocation of public resources like water tankers, garbage trucks, and maintenance teams in real-time.
Zero hardware · 48-hr forecast · Equity-first dispatch
Enter Dashboard →Poor areas served even without complaints
Calculates expected vs actual complaints per ward. When actual < expected, priority is amplified. Systemic under-reporting in low-income wards is corrected to guarantee proportional resource dispatch.
Forecast vs reality time-scrub UI
Side-by-side heatmap layers let operators toggle between prediction and live complaint data. Drag the time slider to scrub through 48-hour windows and verify AI accuracy.
Urgency, emotion & category from 311 text
Fine-tuned BERT model classifies incoming 311 service requests by urgency, location, and service type. "Road collapsed" is prioritised; "grass is long" is not.
Twitter & Reddit fill silent reporting gaps
Mines geo-tagged posts on Twitter and Reddit using BERT classification. Detects "flood here", "garbage piled up" and other hidden problems where formal 311 reporting is absent.
48-hour calendar-aware pre-positioning
Combines historical demand, calendar events (festivals, elections, matches) and weather signals to forecast surge demand 48 hours ahead for proactive fleet staging.
Auto fleet reconfiguration on weather triggers
State machine (Clear→Alert→Warning→Emergency→Recovery) autonomously reconfigures fleet, avoids risky roads, and pre-deploys resources based on NOAA weather feeds — no human needed.
What-if sandbox before committing resources
Full discrete-event simulation. Operators run scenarios — "What if demand +40%?", "What if trucks break?" — and see outcomes on a live map before making real-world decisions.
Garbage + Water + Maintenance on one board
Graph-based conflict detection identifies resource overlaps between sanitation, water, and maintenance departments. Automatically negotiates priority and prevents duplicate routing.
PPO agent: max coverage, min fuel, min time
Proximal Policy Optimization agent trained on historical dispatch scenarios. Suggests which truck goes where with full reasoning. Operators can accept or override every suggestion.
End-of-day LLM-generated KPI PDF
AI pipeline compiles zone coverage, missed deployments, equity scores, prediction accuracy, and operator decisions into a structured PDF daily report with charts and recommendations.
Seven-stage urban intelligence pipeline — raw data to live dispatch