What I Built

carbondebt is a cloud emissions dashboard for backend developers. Put in your cloud provider, region, and monthly compute hours. It calculates your CO2e footprint, breaks it down by service, and uses Google Gemini to return ranked actions with estimated carbon savings per month.

No account. No setup. Results in under ten seconds.

Developers track technical debt across quarters. We argue about it in code review, carry it into sprints. But there is another kind of debt building up in every infrastructure decision, one that never shows up in your AWS bill. Carbon debt works the same way. It compounds quietly and stays invisible until someone surfaces it. This tool surfaces it.

Code

dax-side / carbondebt

carbondebt

carbondebt is a cloud emissions dashboard for developers Enter your workload numbers, then get a monthly footprint estimate and a ranked action list.

Live demo: carbondebt-client.vercel.app

What you get

Total monthly CO2e
Grid intensity for the selected region
Emissions split by service (compute, database, storage, network, lambda)
Best lower-carbon region inside the same cloud provider
Gemini suggestions sorted by estimated monthly savings

How carbon is calculated

The backend uses scope 2 location-based emissions from local region intensity data. Intensity values come from Ember Global Electricity Review 2025.

Formula:

CO2e (kg) = kWh * grid_intensity (gCO2e/kWh) / 1000

Source file for region values: server/src/data/regions.ts

Stack

Frontend: React, Vite, TypeScript, Tailwind CSS
Backend: Node.js, Express, TypeScript
Validation: Zod
AI: Google Gemini
Logging: Morgan + Winston

API

GET /api/health

Returns service status.

POST /api/carbon

Request body:

{
  "provider": "aws"
  "region": "us-east-1",
  "computeHours": 720

…

View on GitHub

How I Built It

The frontend is React with TypeScript and Tailwind. The backend is Node.js and Express. No database. Every request is self-contained.

Carbon calculations run locally on the server using regional grid intensity data from the Ember Global Electricity Review 2025 dataset. No third-party emissions API. The numbers are transparent and traceable to the source.

One thing I added that changed how the data reads: the "equivalent to X km driven" metric. Saying a workload produces 125 kg CO2e per month is accurate but abstract. Converting it to kilometres driven makes the number land for people who have never thought about this before.

Gemini integration: After the carbon report is calculated, it goes to Gemini which returns 3 to 5 ranked infrastructure suggestions specific to the submitted workload. Not generic advice. Things like "switch to Graviton3 instances" or "stop non-production servers outside business hours" with an estimated kg saving attached to each one. The formula gives you the number. Gemini tells you what to do with it.

Prize Categories

Best Use of Google Gemini

Gemini takes the carbon report and turns it into something a developer can act on today. Each suggestion is ranked by estimated monthly CO2e saving and tied to the actual workload submitted. If Gemini is unavailable, the carbon report still renders. The suggestions are an addition, not a dependency.

DEV Community

I built a tool that shows developers what their backend costs the planet