🛡️ GigShield — AI-Powered Parametric Income Insurance for Food Delivery Partners

Guidewire DEVTrails 2026 | University Hackathon Submission Protecting the livelihoods of Zomato & Swiggy delivery partners from uncontrollable external disruptions.

---

📌 The Problem

India's food delivery partners (Zomato, Swiggy) earn ₹15,000–₹25,000/month working outdoors on two-wheelers. When external disruptions like heavy rain, dense fog, or civil unrest hit, platforms reduce order availability or workers are forced to stop — causing 20–30% income loss in a single week with zero financial safety net.

GigShield solves this with a parametric insurance model: no claim forms, no waiting — just automatic payouts when verified disruptions cross defined thresholds.

---

👤 Persona: Food Delivery Partner (Zomato / Swiggy)

User Profile

• Name: Raju, 26 | Operates in Chennai (Anna Nagar + Velachery Zone)
• Avg weekly earnings: ₹4,000–₹6,000
• Working hours: 10 AM – 10 PM, ~6 days/week
• Tech comfort: Moderate — uses smartphone daily for delivery app
• Pain point: No savings buffer; one bad week = skipped EMI or missed rent

Persona-Based Scenarios

Scenario	Disruption	Impact	GigShield Response
Mumbai monsoon week	Rainfall > 50mm/day for 3+ hours	Orders drop 70%, can't ride safely	Auto-trigger: ₹500–₹1,200 payout
Delhi winter fog	Visibility < 50m for 4+ hours	Night deliveries impossible	Auto-trigger: ₹300–₹800 payout
City bandh / protest	Verified civil disruption in worker's zone	Pickup/drop zones blocked	Auto-trigger: ₹400–₹1,000 payout

---

⚙️ Application Workflow

[Worker Onboarding]
        │
        ▼
[Risk Profile Created] ← City, Zone, Avg. Weekly Earnings, Work Hours
        │
        ▼
[Weekly Policy Purchased] ← Dynamic premium shown, UPI payment
        │
        ▼
[Real-Time Monitoring] ← Weather API + News/Alert API polling every 30 min
        │
        ▼
[Disruption Detected] ← Threshold crossed in worker's registered zone
        │
        ▼
[Fraud Check] ← Location validation + anomaly scoring
        │
        ▼
[Claim Auto-Approved] ← Zero manual steps for worker
        │
        ▼
[Instant Payout] ← UPI / wallet transfer within minutes
        │
        ▼
[Worker Dashboard Updated] ← Earnings protected, claim history shown

---

🌧️ Parametric Triggers

GigShield covers income loss only — no health, vehicle, or accident coverage.

Trigger 1: Heavy Rain / Floods

• Data Source: OpenWeatherMap API (free tier)
• Threshold: Rainfall ≥ 40mm in a 3-hour window OR IMD red/orange alert in worker's city
• Payout Logic: ₹500 base + ₹100 per additional disrupted hour (capped at ₹1,200/day)

Trigger 2: Extreme Heat

• Data Source: OpenWeatherMap temperature field
• Threshold: Temperature > 42°C sustained for ≥ 3 hours between 10 AM–4 PM
• Payout Logic: ₹300 base + ₹75 per disrupted hour (capped at ₹800/day)

Trigger 3: Civil Disruption (Protest / Bandh / Curfew)

• Data Source: GDELT Project API + NewsAPI + Twitter/X disaster monitoring feeds
• Threshold: Verified bandh/curfew notice OR 3+ credible news sources reporting civil disruption in worker's registered zone within 1 hour
• Payout Logic: ₹400 flat per disrupted half-day (max ₹1,000/day)

⚠️ Note: All triggers are verified against the worker's registered GPS zone at policy activation time. Claims filed outside the zone are flagged for review.

---

💰 Weekly Premium Model

Gig workers operate and earn on a week-to-week cycle, so GigShield is priced weekly — not monthly or annually.

Base Weekly Premium Tiers

Pricing Rationale: An average Chennai food delivery partner earns ₹4,000–₹6,000/week. Industry best practice keeps insurance premium under 1–1.5% of insured income — making ₹29–₹79/week both affordable and actuarially viable.

Plan	Weekly Premium	Max Weekly Payout	Best For
Basic Shield	₹29/week	₹1,500	Part-time workers (<30 hrs/week)
Standard Shield	₹49/week	₹3,000	Full-time workers
Pro Shield	₹79/week	₹5,000	High-earning / peak-season workers

AI-Driven Dynamic Pricing Adjustments

The base premium is adjusted weekly using ML risk factors:

Risk Factor	Adjustment
Zone historically flood-prone (e.g., low-lying areas)	+₹5–₹15/week
Worker's city has IMD pre-season warning	+₹10/week
Worker's zone historically low disruption	−₹5/week
Worker has 0 claims in last 4 weeks	−₹3/week (loyalty discount)
Predicted rain probability next 7 days > 70%	+₹8/week

Premium is recalculated and shown to the worker every Sunday before the new week begins. Worker must actively renew — no auto-debit surprise charges.

---

🤖 AI/ML Integration Plan

1. Dynamic Premium Calculation (Risk Scoring Engine)

• Model: Gradient Boosted Trees (XGBoost / LightGBM)
• Inputs: Zone flood history, AQI trends, seasonal weather patterns, worker's claim history, city-level disruption frequency
• Output: Risk score (0–100) → maps to weekly premium adjustment
• Phase 1: Rule-based mock; Phase 2: trained on synthetic + public weather data

2. Fraud Detection Engine

• Anomaly Detection: Isolation Forest model
• Signals monitored:
  • GPS location mismatch (worker not in registered zone during claimed disruption)
  • Claim filed despite active delivery records on platform (simulated Swiggy/Zomato activity feed — if worker is actively completing orders, disruption claim is flagged)
  • Multiple claims in rapid succession
  • Device fingerprint inconsistency
• Output: Fraud risk score → Auto-approve (low), Flag for review (medium), Reject (high)

3. Predictive Disruption Alerts

• Model: Time-series forecasting (Prophet / LSTM) on weather data
• Use: Pre-warn workers of likely disruption next day; allow insurers to provision payout reserves
• Phase 3 feature

4. AI Risk Map (Visual Analytics Dashboard)

• What it shows: Chennai zone-level risk heatmap
  • 🔴 High flood/heat risk zones (e.g., Velachery, Tambaram)
  • 🟡 Moderate risk zones
  • 🟢 Low risk / safe zones
• Benefits: Drives hyper-local premium pricing; gives insurers real-time portfolio risk view
• Tech: Google Maps SDK + historical weather + claim data overlay

---

🛠️ Tech Stack

Mobile App (Frontend)

• Framework: React Native (Expo) — cross-platform iOS + Android
• UI Library: React Native Paper / NativeWind (Tailwind for RN)
• State Management: Zustand
• Maps: React Native Maps (Google Maps SDK)

Backend

• Runtime: Node.js + Express.js
• Database: PostgreSQL (user profiles, policies, claims) + Redis (real-time trigger cache)
• Auth: Firebase Auth (phone number OTP — familiar to gig workers)
• Job Scheduler: Bull Queue (for periodic weather API polling)

AI/ML

• Language: Python (FastAPI microservice)
• Libraries: Scikit-learn, XGBoost, Prophet
• Serving: REST API called by Node backend

Integrations

Integration	Provider	Mode
Weather data	OpenWeatherMap API	Real (free tier)
Civil disruption alerts	Google Alerts RSS + admin panel	Mock (Phase 1-2)
Payment gateway	Razorpay Test Mode	Sandbox
Platform activity (Zomato/Swiggy)	Simulated delivery data	Mock

Infrastructure

• Hosting: Railway.app / Render (backend) + Expo EAS (mobile builds)
• CI/CD: GitHub Actions
• Version Control: GitHub (this repo)

🏗️ Repository Structure (Planned)

gigshield/
├── mobile/               # React Native app
│   ├── screens/
│   ├── components/
│   └── navigation/
├── backend/              # Node.js + Express API
│   ├── routes/
│   ├── services/
│   │   ├── weatherService.js
│   │   ├── triggerEngine.js
│   │   └── fraudService.js
│   └── models/
├── ml/                   # Python FastAPI ML service
│   ├── premium_model/
│   └── fraud_model/
├── docs/                 # Architecture diagrams, wireframes
└── README.md

---