Pabandi Portal: AI case management for Uttar Pradesh Police

The challenge

Police stations in Chandauli district ran preventive-action cases (BNSS Sections 126 and 135) entirely on paper. Case documents arrived unstructured, in Hindi, English, and transliterated Hindi. Legal notices moved through an eight-stage approval workflow tracked by hand, and with no shared system across the district's stations, spotting a repeat offender who surfaced at a different station was close to impossible. Officers spent hours on administration that should have gone to policing.

What we built

An end-to-end AI case management platform:

• Multilingual AI document extraction — GPT-4o with custom prompting and confidence scoring reads Hindi and English case documents at 95%+ accuracy, turning a 20-minute manual process into one that takes under 30 seconds.
• Risk scoring and workflow tracking — every case is automatically classified high/medium/low risk, with real-time visibility across the eight-stage approval workflow.
• Cross-station repeat-offender detection — instant record lookup across every connected station in the district.
• Instant search — full-text and semantic search across 7,849+ documents, answering in under 2 seconds.
• Role-based access control — station, supervisor, and district-level data segregation.

Under the hood: React 19 and TypeScript on the front end, Node.js and PostgreSQL behind it, deployed on AWS with Docker and CI/CD.

The outcome

• 70% reduction in administrative work per officer — 10+ hours saved per officer, per week
• 5,000+ active preventive cases under review, 26,000+ records tracked
• 18+ police stations connected and live

The deployment was covered by Hindustan Times, with the district Superintendent of Police describing how disputes recorded at any station are now uploaded to the portal immediately.

The service behind it

This is what our AI-Native Development service looks like in production: AI is the core of the system, not a feature bolted on — and it works in Hindi as well as English.