KSP Crime Genome — AI-Powered Investigation OS

Winner — Karnataka State Police Datathon 2026

One platform that takes a crime from FIR to chargesheet — extracting entities in Kannada, cross-matching evidence across cases, scoring readiness, and tracking every legal deadline.

The Problem

Karnataka Police officers manage cases across 27+ CCTNS tables, disconnected spreadsheets, and paper trails. Evidence gets buried, cross-FIR connections go unnoticed, chargesheet deadlines slip, and critical gaps surface only when a judge asks. Officers spend more time hunting for information than analyzing it.

The Solution

Crime Genome unifies the entire investigation lifecycle into one AI-powered operating picture. Enter an FIR number, and the system instantly ingests the case, extracts entities via ZIA Copilot (Kannada + English), cross-references evidence against 1,200+ synthetic cases, scores chargesheet readiness, and tracks every legal deadline — all in a single view.

Demo

Live site: http://localhost:5173 (or deployed URL)

Username	Password
anjumala	123

Click "Enter the Command Center" on the landing page for a one-click ACP demo session.

Key Features

• ZIA Copilot — Kannada/English AI assistant powered by Sarvam AI, with voice support
• Crime Hotspot Prediction — XGBoost model trained on Karnataka crime data for spatial forecasting
• Evidence Pipeline — FIR → entities → cross-match → chargesheet in one view
• Face Analytics — Age/gender/emotion detection from suspect photos
• Object Recognition — 10-class police evidence detection pipeline
• Text Analytics — Sentiment + entity extraction from witness statements
• Chargesheet Clock — Countdown to every legal deadline with urgency alerts
• Entity Graph — Cross-FIR connection visualization with force-directed layout
• Live Demo Mode — ?live=true for real-time crime simulation
• Role-based Dashboards — ACP, Inspector, Sub-Inspector, Supervisor views
• Person 360 — Entity risk profiling with cross-case timeline
• Theory Board — Classify and track investigative theories
• Beat Optimizer — ML-driven patrol route optimization
• Dark Figure Estimation — Unreported crime statistical modeling

Architecture

┌─────────────────────────────────────────────────────┐
│                   React + Vite                       │
│         (Slate frontend, 86 components)              │
├─────────────────────────────────────────────────────┤
│              27 Catalyst Serverless Functions         │
│  ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────────┐  │
│  │ Case  │ │ Chat │ │ ZIA  │ │ Co-  │ │ Chargesheet│  │
│  │ Mgmt  │ │Query │ │Brief │ │Accused│ │  Clock   │  │
│  └──────┘ └──────┘ └──────┘ └──────┘ └──────────┘  │
├─────────────────────────────────────────────────────┤
│             26/26 Catalyst Services Utilized          │
│     Data Store · Cache · Email · Queue · Cron · ... │
├─────────────────────────────────────────────────────┤
│  Sarvam AI  │  XGBoost  │  Leaflet  │  Force-Graph  │
│  (Kannada   │  (Hotspot │  (Maps)   │  (Entity      │
│   LLM+TTS)  │  Pred.)   │           │   Network)    │
└─────────────────────────────────────────────────────┘

• 27 Catalyst serverless functions — case management, crime chat, ZIA brief, co-accused network, chargesheet clock, face analytics, object recognition, text analytics
• React + Vite frontend on Catalyst Slate
• XGBoost + spatial ML for hotspot prediction
• Sarvam AI for Kannada LLM, speech-to-text, and text-to-speech
• 26/26 Catalyst services utilized — full platform coverage

How It Works

1. Enter an FIR number (e.g., KSP-2026-0142) → system loads case details
2. ZIA extracts entities (suspects, vehicles, phones, locations) in Kannada + English
3. Evidence cross-matched against linked cases with similarity scoring
4. Evidence completeness scored → gaps flagged with actionable warnings
5. Chargesheet readiness calculated → legal deadline tracked with countdown

Demo Script (for judges)

1. Go to the live URL
2. Click "Enter the Command Center" — auto-login as ACP Demo Officer
3. You are at the DemoHub. Type KSP-2026-0142 in the search box
4. Watch the 4-stage pipeline animate (FIR → Entities → Evidence → Chargesheet)
5. Click "Ask ZIA" — try "what evidence is missing?" (Kannada supported)
6. Click "View on Map" — see the crime hotspot prediction
7. Add ?live=true to the URL — see real-time crime alert simulation
8. Try alternate FIRs: KSP-2026-0089, KSP-2026-0201

Team

Rushdarshan — [Your contact]

Built With

Technology	Purpose
Catalyst by Zoho	Serverless backend + 26 services
Sarvam AI	Kannada LLM, STT/TTS
XGBoost	Crime hotspot prediction
React + Vite	Frontend framework
GSAP	Landing page animations
Leaflet	Crime mapping
Force Graph	Entity network visualization
ApexCharts	Analytics dashboards
Radix UI	Accessible components
Python	Synthetic data generation, ML pipeline