EcoSense

Inspiration

Climate impact is something we hear about constantly, but it’s often hard to translate into everyday action. We wanted to build something that makes emissions feel tangible and personal rather than abstract. EcoSense was inspired by the idea of giving users a simple way to log real-life activities—like transportation, electricity use, and food consumption—and immediately see their environmental impact. We also wanted to create something that doesn’t just track data, but actually helps users understand it over time through visualization and AI-driven feedback.

What it does

EcoSense is a carbon footprint tracking web application that helps users measure, visualize, and understand their environmental impact.

Users can sign up, log in, and securely store their emission-related activity data
Tracks transportation (distance-based travel), home energy usage (kWh-based), and food consumption
Uses predefined emission factors stored in the database to calculate consistent CO₂ estimates
Displays a dashboard with totals, breakdowns, and visual charts
Includes an AI Eco Advisor powered by Google Gemini that provides personalized insights based on user history
Allows users to export a PDF report summarizing their footprint and AI-generated insights

In short: log → analyze → visualize → understand → export

How we built it

EcoSense is a full-stack web application built with a React + TypeScript frontend and a Node.js + Express backend.

Frontend

React 19 with Vite
Context API for authentication and emissions state
CSS Modules for styling
Client-side routing for navigation

Backend

Express.js with TypeScript
PostgreSQL database
Sequelize ORM with migrations and seeders
JWT authentication and protected routes
Joi validation for request safety

AI Integration

Google Gemini API (@google/generative-ai)
System prompts include structured user footprint data when available
Context-aware responses based on logged emissions

Other Tools

jsPDF for generating downloadable reports
REST API structure under /api/v1/...
CORS configured for secure frontend-backend communication

We model emissions using:

[ E = A \cdot f ]

where (E) is emissions (kg CO₂e), (A) is activity intensity, and (f) is the emission factor.

Total emissions:

[ E_{\text{total}} = \sum_i E_i ]

Challenges we ran into

Keeping frontend, backend, and database models consistent across multiple emission categories
Managing authentication flows (JWT persistence and /me session refresh)
Passing structured user data into the AI without exceeding token limits
Ensuring the AI advisor only personalizes responses when data exists
Making PDF export reliable even when API or AI responses are slow
Debugging migrations, seeders, and CORS issues during integration

Accomplishments that we're proud of

Fully working end-to-end system: auth, logging, dashboard, AI advisor, and PDF export
Centralized emission factors stored in the database (not hardcoded)
AI advisor that meaningfully adapts to user-specific footprint data
Clean separation between frontend and backend services
A complete product that turns abstract emissions into actionable insights

What we learned

How to design APIs around real user workflows (log → store → analyze → visualize)
Sequelize migrations and seeders as a reliable source of truth
Effective prompt and context design for LLM integration
Trade-offs between client-side vs server-side API key handling
Full-stack debugging across frontend, backend, and database layers