Hackathon IBM x ESILV
As the use of generative AI continues to grow, its environmental impact has become a key concern.
Track Sustainability provides a solution to measure and visualize the carbon footprint of a query executed on a Large Language Model (LLM).
The goal is to raise awareness among users about the energy consumption and CO₂ emissions associated with AI usage, while offering clear and relatable comparisons (e.g., equivalent kilometers by car, duration of lighting a bulb, etc.).
We are developing an interactive web application that looks just like a standard LLM chat interface.
When a user submits a request:
- The request is processed by a pre-trained model capable of estimating the energy cost of handling the query.
- This estimated cost is converted into CO₂ emissions.
- The app displays a visual comparison to help the user understand the impact of their query.
The project leverages the WatsonX platform for model interaction, computation, and performance evaluation.
| Technology / Tool | Purpose |
|---|---|
| WatsonX | Runtime environment and model analysis |
| Machine Learning / Deep Learning | Estimation of computational and environmental cost |
| Web Application (Frontend + Backend) | User interface and application logic |
| Data Visualization Tools | Clear and intuitive representation of CO₂ impact |
Project developed as part of the IBM x ESILV Hackathon by:
- Paul Ranc
- Héloïse Roméo
- Natalia Gerard
- Alexis Denneulin
- Valentin Rech
- Valentin Templé
Our ambition is to create a reference tool for environmental awareness in AI usage.
Understanding the impact is the first step toward responsible innovation.