By Team Code Alchemists
KL University | DEVTrails 2026
🚀 Phase 1 vs. Phase 2: From Prototype to Product
Phase 1 proved the concept; Phase 2 delivers a market-ready infrastructure.
Feature Phase 1 (Seed) Phase 2 (Scale)
Platform React Web Prototype Native Flutter App (45+ Screens)
Backend Mock Data (No Database) Real Firebase + Cloud Functions
Auth Manual Entry Phone OTP + Biometric Fingerprint
Verification Simple GPS check 4-Layer AI Evidence Verification
Policy Static Plans Dynamic Premium (Zone + Platform risk)
Support None FAQ Bot + 24hr Ticket System
🧠 The "Zero-Touch" Claims Flow
The core of GigWeatherWage is its ability to pay workers without them filing a single piece of paperwork.
code:
graph LR
A[Weather API] -->|Rain Detected| B(Push Notification)
B --> C{Worker Taps 'Yes'}
C --> D[5-Signal AI Fraud Check]
D -->|Verified| E[Decision < 30 Seconds]
E --> F[Instant UPI Payout]
📊 Dynamic Premium Logic
GigWeatherWage doesn't believe in "one size fits all." Premiums are calculated every week based on specific risk factors.
Zone Risk Multipliers (High vs. Low Risk)
A visual representation of how geography affects the cost of insurance:
City - Area Risk Level Multiplier
Mumbai - Dharavi Very High 1.5x
Hyderabad - Madhapur High Flood 1.4x
Chennai - Anna Nagar High Heat 1.3x
Bengaluru - Koramangala Low Risk 0.9x
Platform Risk Factors
Zepto (10-min delivery) riders face higher risks than Amazon (long window) delivery partners:
Zepto: 1.2x 🟥🟥🟥🟥🟥 (High Pressure)
Swiggy: 1.1x 🟥🟥🟥🟥⬜ (Peak Pressure)
Zomato: 1.0x 🟥🟥🟥⬜⬜ (Standard)
Amazon: 0.9x 🟥🟥⬜⬜⬜ (Long Windows)
🛡️ The 4-Layer AI "Fraud Shield"
To prevent fake claims, every photo evidence must pass through four distinct AI checkpoints:
L1: AI Image Detection: Scores 0-100 to detect GAN/Midjourney generated fakes. (Reject if >70).
L2: Reverse Image Search: Checks Google Vision to ensure the photo isn't from the web.
L3: Face Match + Liveness: Compares a live selfie with stored face embeddings. Blink detection prevents "photo-of-a-photo" fraud.
L4: Metadata + Timestamp: Ensures the photo was taken within 30 mins and 500m of the claim.
💰 The Economic Engine (Money Flow)
How the platform sustains itself while paying out claims.
Worker Pays: Rs. 20-80/week (The Premium).
Risk Pool: All premiums flow into a collective pool.
Underwriter: Large insurance partners back the pool to cover major shortfalls.
GWW Fee: GigWeatherWage earns a 15-20% platform fee for operations and AI maintenance.
👥 Meet the Personas (Real Data Demo)
The system is tested against four distinct user profiles to ensure fairness and security:
Persona Status Risk Score Outcome
Raju Kumar Genuine Worker 12 Paid Instantly
Meena Devi New Account 55 Delayed 2 Hours
Priya Sharma Insider Fraud 85 Blocked
Vikram #7749 GPS Spoof Ring 135 Blocked + Ring Added to DB
📍 Project Roadmap
Phase 1 (SEED) - [COMPLETED]: Architecture design, React prototype, 5-signal fraud engine.
Phase 2 (SCALE) - [CURRENT]: Flutter App, Firebase Backend, Aadhaar & Face Auth, Admin Panel.
Phase 3 (SOAR) - [PLANNED]: Pilot with 10 real workers in Hyderabad, IRDAI license application, platform API integrations.
Top comments (0)