Eco Score

Logo
Home Page
Energy
Transportation
Rewards
Dashboard
Categories
Shopping

Inspiration

Our vision for this app is to transform the way people understand and interact with their personal carbon footprint. Just as a credit score has become a universally recognized measure of financial responsibility, we aim for the carbon emission score to become an equally standard metric for environmental responsibility.

By providing users with clear, actionable insights into the emissions generated by their daily choices—whether it’s transportation, energy use, or shopping—we empower individuals to make smarter, greener decisions.

This app isn’t just a tool; it’s a movement toward mainstream environmental accountability, helping people see the impact of their actions in real time and inspiring a global shift toward a more sustainable future.

What it does

EcoScore is a comprehensive mobile platform that tracks carbon footprint across multiple lifestyle categories like shopping list, transportation, energy consuption and converts all actions into a universal score redeemable for real cashback, discounts, and rewards at participating stores.

The app places users in a percentile ranking, showing how their impact compares to the broader community. To make sustainable living even more rewarding, users in higher percentile tiers unlock discounts and special offers at major stores, creating a powerful incentive to make smarter, greener choices every day.

How we built it

Mobile App: React Native + Expo. Users snap or upload a photo of their shopping or utility bill, which we POST as multipart to FastAPI, then render results in a clean UI. The EcoScore ring + tiers make it feel like a “score” you want to improve.

Backend: Python + FastAPI. One service handles uploads (receipts, energy bills, transport) and returns clean JSON responses.

OCR Pipeline: Google Cloud Vision (for images) and pdfminer.six (for text-based PDFs). Regex and heuristics clean and extract items, totals, currencies, and transport details.

LLM Scoring: Parsed receipt items go to OpenAI (gpt-4o-mini) in batch to estimate kg CO₂e per item. Prompts are optimized to return concise JSON: item name + emissions.

Data Layer: MongoDB for receipts, energy, transport, and a simple points ledger.

Points + Percentile: Every action becomes points. We aggregate today’s entries and compute percentile rankings across all users. No big data stack—just math over a flat list.

Challenges we ran into

1. OCR Accuracy and Text Parsing Complexity

Challenge: Different receipt formats, energy bills, and transportation receipts vary widely. OCR often produced noisy, fragmented output. Solution: Implemented sophisticated regex-based parsing with multiple patterns to handle:

Currency detection (USD, EUR, GBP, INR)
Amount extraction with OCR error correction (e.g., “30 83” → “30.83”)
Date parsing (MM/DD/YYYY, Month DD, YYYY)
Distance extraction for transportation
Energy bill parsing (kWh, billing periods, TOU rates)

2. Points Calculation System

Challenge: Designing a fair scoring system that rewards low-carbon choices across categories. Solution:

Shopping: points = max(0, 10 - carbon_emission)
Energy: points = 100 - carbon_emission
Transportation: points = max(0, 10 - carbon_emission) Unified daily/monthly aggregation ensures consistency.

3. Google OAuth and Authentication

Challenge: Secure Google OAuth flow with JWT generation and Gmail API prep. Solution:

Google ID token verification using google-auth
JWT generation with 7-day expiry
MongoDB user storage
Graceful handling of missing Gmail scopes

4. Data Persistence and Structure

Challenge: Managing multiple JSON files (receipts.json, energy.json, transport.json, points.json) while maintaining consistency. Solution:

Unified, user-based data model
UUID-based entry IDs
Atomic file operations
Flat points.json for efficient queries

Accomplishments that we're proud of

1. Comprehensive Carbon Tracking System

Full-stack app covering Shopping, Energy, and Transportation
Real-time carbon calculations using standard emission factors
Unified gamified rewards system

2. Advanced OCR and Text Processing

Robust OCR pipeline with Google Cloud Vision
Smart text parsing for:
- Item extraction (quantity + price)
- Energy bill parsing (kWh, TOU, supplier info)
- Transport data (distance, duration, route)
OCR error correction for fragmented values

3. AI-Powered Carbon Estimation

Integrated GPT-4 for emission estimation
Batch processing for efficiency
Fallback handling for ambiguous products
Structured JSON responses with confidence scores

What we learned

1. OCR and Text Processing

Regex normalization of OCR output
Handling edge cases (split values, multiple currencies, date formats)
Experience with Google Vision API
Extracting structured data from unstructured text

2. LLM Integration and Prompt Engineering

Crafting stable prompts for JSON outputs
Batch optimization for cost
Handling ambiguity with context
Temperature tuning for reliability

3. Carbon Footprint Calculation

Emission factors across activities
Market vs. location-based accounting
Vehicle-specific emissions
Energy conversion tracking

4. Project Management

Breaking complex features into sprints
Prioritizing MVP delivery
Managing technical debt
Team collaboration and version control

What's next for Eco Score

Vision

Right now, users enter their activities manually, but in the future our vision is that this data will be tracked automatically, with partner companies securely sending us each user’s activity directly from their systems.