This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
I built ExoSight AI Explorer for a friend who loves space but doesn’t have a background in astronomy or machine learning. The goal was to make NASA exoplanet data feel accessible, visual, and exciting instead of overwhelming.
ExoSight AI Explorer is an AI-powered web app that analyzes NASA Kepler/TESS exoplanet data, trains a machine learning model to identify promising exoplanet candidates, and presents the results in a simple interactive dashboard. It helps users explore candidate planets, review key features like orbital period, planetary radius, temperature, and insolation, and understand which objects stand out as high-potential exoplanets.
This project turns raw scientific data into a more intuitive experience for curious people who want to explore the universe without needing to parse complicated research datasets.
Demo
This project is designed to run locally as a Streamlit app:
- Launch with:
streamlit run app.py - The repo includes the trained model, data processing scripts, and a web interface for exploring exoplanet candidates.
While I don’t have a public deployment link yet, the full app and source code are available in the repository below.
Code
https://github.com/SnehaghoshBarsha444/nasa-space-apps-exosight-ai-explorer
How I Built It
I built ExoSight AI Explorer using open-source tools and public astronomy data.
- Data source: NASA exoplanet datasets from Kepler/TESS-style catalog data
- Frontend/app interface: Streamlit
- Data processing: pandas
- Visualization: Plotly
- Model training: scikit-learn
- Model persistence: joblib
The app processes exoplanet features like:
- orbital period
- planetary radius
- equilibrium temperature
- duration
- impact parameter
- insolation
It then uses a Multi-Layer Perceptron (MLP) classifier to flag exoplanet candidates with high confidence. The model identifies likely candidates from the dataset and stores them in a generated CSV for exploration in the app. The result is an approachable interface that makes it easier to browse candidate exoplanets instead of staring at raw scientific tables.
Why Does Open Innovation Matter?
Open innovation matters here because the foundation of this project is built on public NASA data and open-source software. Without open datasets, transparent model tooling, and community-driven scientific tooling, this kind of project wouldn’t be possible in such a lightweight and accessible way.
Using open-source libraries like pandas, scikit-learn, and Streamlit made it possible to:
- train an ML model on real astronomy data
- build a user-friendly interface quickly
- make the project easy to reproduce and extend
- share results publicly with the broader community
A closed API or proprietary system would have made this much harder to build in a transparent, explainable, and reproducible way. Open innovation lets people explore science together, not just consume it.
My Agent Session
Optional, but not included here.
Prize Categories
- Open Source AI / Open-Weight Models
- Space + Science
- Best Use of AI
Top comments (0)