Project 03 · Hackathon build
FuelGuard
FuelGuard - AI-Powered Fuel Fraud Detection
Don’t get cheated at the pump.
FuelGuard is a full-stack web app that uses AI to detect fuel pump fraud from uploaded videos, generate legal PDF reports, and map fraudulent pumps in real time.
- Type
- Hackathon build
- Built at
- Robotrix 2026
- Core stack
- React 18 · TypeScript · Vite
- Live at
- fuel-guard-dev.vercel.app
Overview
What it is
FuelGuard uses computer vision to catch fuel pump fraud. You upload a video of your fueling session, the AI service reads the meter frame by frame, and the app tells you if something is wrong.
Every analysis can become a PDF evidence report and a point on an interactive fraud heatmap. Dedicated portals give fleet operators and government inspectors their own view of the same data.
Architecture
How it’s built
- 01React + TypeScript appUploads, reports, heatmap and portals
- 02Express APIAuth, payments and PDF generation
- 03FastAPI + OpenCVReads the meter frame by frame
- 04MongoDB + CloudinaryRecords and video storage
Features
What it does
For drivers
- Video analysisUpload a fueling video for AI fraud detection.
- PDF evidence reportsGenerate evidence documents from an analysis.
- Fraud heatmapSee fraud hotspots across cities on an interactive map.
- Pump reports and reviewsBrowse fraud scores and community ratings for pumps.
- English and HindiMulti-language support for a wider audience.
Fleet management
- Vehicle trackingMonitor fuel consumption across a fleet.
- Fill historyEvery fill logged with its location.
- Fraud flaggingSuspicious fills are flagged automatically.
- Monthly reportsPDF reports covering fleet expenses and fraud.
Government portal
- Live dashboardOverview of fraud reports and inspections.
- City analyticsFraud broken down by city and fuel company.
- Priority alertsFlags pumps with high fraud rates for inspection.
Tech stack
Built with
Frontend
- React 18
- TypeScript
- Vite
- Tailwind CSS
- Framer Motion
- Recharts
- Leaflet
Backend
- Node.js
- Express
- MongoDB
- Mongoose
- JWT
- Cloudinary
- Razorpay
- PDFKit
- Ethers.js
AI service
- Python
- FastAPI
- OpenCV
Screens
A closer look
Built at a hackathon
Robotrix 2026
Participant
- Organiser
- NITK Surathkal
- Platform
- Unstop
- Date
- Jan 2026
- Format
- Team event
- Team
- 3 members
- Focus
- Web · Mobile · AI
- Our build
- FuelGuard - AI-Powered Fuel Fraud Detection