Skip to content
Projects · Apr 2026 · Hackathon

ECOpath

Carbon-aware, agentic multi-modal route planner — hackathon build comparing every way from A to B by carbon, time, and cost, with LLM reasoning over verifiable EPA/IPCC/FTA data.

Agentic AI Sustainability Full-Stack Hackathon

Overview

Most map apps tell you the fastest route. ECOpath shows the greenest, cheapest, and fastest options side-by-side, with a natural-language recommendation that respects real constraints ('be there by 10 AM', 'under $5', 'I care about emissions').

Problem

Carbon-aware routing needs live data, defensible emissions math, and reasoning over user constraints — without hardcoding or hand-waving the numbers.

Approach

A three-stage agentic pipeline with strict boundaries: a Routing Agent fetches live options from the Google Maps API, an Emissions Agent computes physics from EPA/IPCC/FTA datasets, and a Decision Agent (Groq Llama) reasons and explains — no agent reaches around another. Built with Kiro (hooks, steering, specs), FastAPI, Next.js, Supabase auth, and Google Calendar OAuth.

Impact

A complete agentic product built under hackathon time pressure: multi-agent boundaries, verifiable data lineage, and a polished full-stack UX.