"GigShield" — Parametric Income Insurance
An AI-powered parametric income protection platform for gig delivery workers. When verified disruptions like heavy rain, hazardous air quality, or platform outages cut into earning hours, GigShield triggers automatic, instant payouts — no forms, no claims processing. Features an immersive 3D landing experience, location-based triggers, and a Supabase-backed Express API.
- My role
- Team of three — I built the trigger engine, payout lifecycle and receipts, the rider dashboard, admin and fraud-review screens, role-based routing and the 3D landing; the liveness check with Rishabh
- Built for
- Team hackathon build · Mar–Jun 2026
“For a worker paid weekly, a claim that settles next month is no claim at all.”
The Brief
India’s delivery riders earn weekly on thin cash buffers. One day of heavy rain, a hazardous AQI reading or a platform outage can wipe out a day or two of income — and traditional insurance, built around documents, adjusters and reimbursement weeks later, doesn’t fit a loss that small and that sudden.
Parametric cover fits better: pay a set amount when an objective, external signal crosses a threshold. The hard part is trusting the signal and stopping the same disruption from paying out twice.
The Approach
Riders pick a weekly plan (Basic ₹79, Standard ₹129, Pro ₹179) with a daily payout cap and covered hours. A trigger engine watches for disruptions — heavy rain, heatwave, AQI spikes, platform outages, curfews, strikes and zone closures — and scores each with a confidence value built from a per-trigger base, weather-source reliability and the rider’s risk level.
Weather is read from Open-Meteo with an automatic fallback to wttr.in, and confidence rises when more sources agree. Cooldown and de-duplication windows stop one event from firing repeatedly, every trigger is written to an audit trail synced to the backend, and claims move through a visible lifecycle — pending verification → verified → processing → settled — with fraud checks and a liveness check on risky sessions. That check asks for a random hand gesture, which MediaPipe’s hand-landmark model classifies right in the browser.
How It Works
- 01Weather & statusOpen-Meteo, wttr.in fallback
- 02Trigger engineconfidence, cooldown, de-dup
- 03Fraud checksanomaly review, liveness gesture
- 04Payout lifecyclepending → settled
- 05Dashboardsrider + admin views
Key Decisions
Weekly, not monthly
Premiums, caps and covered hours are all weekly, because that’s how riders are paid and how they think about money.
Confidence over a yes/no
Each trigger carries a Strong / Normal / Weak confidence label instead of a bare boolean, so admins can see how much the system trusted the signal behind a payout.
Two weather sources
A 2.5-second timeout on Open-Meteo falls back to wttr.in, so one flaky API can’t silently stop the engine.
Income only
The product deliberately covers lost earning hours and nothing else — no health, life, accident or vehicle cover.
- Payouts and parts of the fraud workflow are simulated — no real payment rail is connected.
- Trigger cooldown and de-duplication state lives in the browser’s storage, not on the server.