AI/ML application for NASA Space Apps Challenge 2025: "A World Away: Hunting for Exoplanets with AI"
An interactive web application that leverages NASA's open exoplanet datasets to analyze space mission data and automatically identify exoplanet candidates using machine learning. Built with cutting-edge technologies and designed for astronomical discovery.
- 🤖 ML-Powered Predictions: Binary classification model (Random Forest) trained on KOI, K2, and TESS mission data
- 🌍 Interactive 3D Visualization: Immersive planet explorer built with Three.js
- 📊 Real-time Analytics: Comprehensive model metrics and performance dashboards
- 🌐 Multilingual Support: Full i18n support (English & Spanish)
- 🎨 Modern UI: Built with Nuxt 4, Vue 3, and Tailwind CSS
- ⚡ Fast API: Python FastAPI backend with optimized predictions
- 📈 Data Pipeline: Automated harmonization of multiple NASA datasets
- Nuxt 4 - Meta-framework for Vue 3
- Vue 3 - Progressive JavaScript framework
- Tailwind CSS - Utility-first CSS framework
- Three.js - 3D graphics library
- shadcn-vue - Beautiful UI components
- TypeScript - Type-safe development
- FastAPI - Modern Python web framework
- scikit-learn - Machine learning library
- pandas - Data manipulation and analysis
- joblib - Model serialization
- Pydantic - Data validation
- Moonrepo - Monorepo orchestrator
- pnpm - Fast, disk space efficient package manager
- uv - Ultra-fast Python package installer
- Docker - Containerization
- ESLint - Code linting
- Ruff - Python linter/formatter
AWorldAway_Nasa2025/
├── apps/
│ ├── web/ # Nuxt 4 frontend application
│ │ ├── app/ # Application code
│ │ ├── i18n/ # Internationalization
│ │ └── components/ # Vue components
│ └── server/ # FastAPI backend
│ ├── src/ # Source code
│ ├── models/ # Trained ML models
│ └── assets/ # Static assets
├── packages/
│ └── analytics/ # Data processing & ML pipeline
│ ├── data/ # Raw and processed datasets
│ ├── scripts/ # Python scripts for data harmonization
│ └── .output/ # Generated models and reports
├── scripts/ # Build and deployment scripts
├── .moon/ # Moonrepo configuration
└── moon.yml # Root task orchestration
1. Install pnpm:
# Via npm
npm install -g pnpm
# Or via Corepack (recommended)
corepack enable
corepack prepare pnpm@latest --activate2. Install uv (Python package manager):
# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
# Windows (PowerShell)
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
# Or via pip
pip install uv3. Install moonrepo:
npm install -g @moonrepo/cli-
Clone the repository
git clone https://github.com/lunajimenez/AWorldAway_Nasa2025.git cd AWorldAway_Nasa2025 -
Install dependencies
pnpm install
-
Run the application 🎉
moon source:dev
This single command will:
- ✅ Generate the ML model from harmonized datasets (
analytics:generate) - ✅ Copy the model to the server (
source:copy-model) - ✅ Start the FastAPI backend on
http://localhost:8000 - ✅ Start the Nuxt frontend on
http://localhost:3000
- ✅ Generate the ML model from harmonized datasets (
-
Open your browser
http://localhost:3000
# Run full development environment (recommended)
moon source:dev
# Run only the web frontend
moon web:dev
# Run only the API server
moon server:run
# Regenerate ML model from datasets
moon analytics:generate
# Copy generated model to server
moon source:copy-model# Prepare production build (generate model + copy)
moon source:prepare
# Deploy (requires configuration)
moon source:deploy# Install web dependencies
moon web:pnpm -- install
# Add shadcn-vue components
moon web:shadcn -- add button
# Run Python scripts with uv
moon analytics:uv -- run scripts/harmonize_intersection.pyOur binary classification model achieves:
- ROC AUC: 88.89%
- PR AUC: 77.74%
- Accuracy: 81.24%
- Macro F1: 78.68%
- Optimal Threshold: 0.3433
The model uses 9 features including:
- Orbital period (days)
- Transit duration (hours)
- Planet radius (Earth radii)
- Equilibrium temperature (K)
- Transit depth (ppm)
- Insolation flux (Earth flux)
- Stellar radius (Solar radii)
- Stellar temperature (K)
- Source mission (Kepler/K2/TESS)
GET /api/model/predict-one- Single exoplanet candidate predictionPOST /api/model/predict-params- Batch predictionsGET /api/model/metrics- Model performance metricsGET /api/model/info- Model configuration
- Interactive API docs:
http://localhost:8000/docs - OpenAPI schema:
http://localhost:8000/openapi.json
Explore a procedurally generated planet with orbiting moons using Three.js. Features include:
- Real-time camera controls (WASD + mouse)
- Dynamic lighting and shadows
- Procedural textures for planets and moons
- Starfield background with 1000+ stars
User-friendly form to input exoplanet parameters and get instant predictions with:
- Real-time validation
- Confidence scores
- Visual result modal with interpretation
- Multilingual support
Comprehensive visualization of model performance:
- Confusion matrix
- Classification report
- ROC/PR AUC scores
- Dataset statistics
- Feature importance
Create a .env file in the apps/server/ directory:
# Server Configuration
HOST=0.0.0.0
PORT=8000
# CORS Origins (comma-separated)
CORS_ORIGINS=http://localhost:3000,http://127.0.0.1:3000For the frontend (apps/web/.env):
API_BASE_URL=http://localhost:8000Build and run with Docker Compose:
moon source:deployOur data processing pipeline:
- Data Collection: KOI, K2, and TESS datasets from NASA
- Harmonization: Align features across different missions using
uvscripts - Feature Engineering: Calculate derived physics-based features
- Model Training: Random Forest classifier with hyperparameter tuning
- Evaluation: Comprehensive metrics and visualization
- Deployment: Model serialization (joblib) and API integration
The entire pipeline is orchestrated by moonrepo and runs with a single command.
StardustCrusaders - Universidad Tecnológica de Bolívar
- Luna Jiménez - Team Leader & Data Scientist
- Leonardo González - Data Scientist & Hardware Developer
- Maria García - Data Scientist & Backend Developer
- Michael Taboada - Data Scientist & Backend Developer
- Mauro Gonzalez - Fullstack Developer & Data Scientist
This project is licensed under the MIT License - see the LICENSE file for details.
- NASA for providing open exoplanet datasets
- Space Apps Challenge organizers
- The open-source community
- Astral for the amazing
uvtool - Moonrepo for monorepo orchestration
- NASA Exoplanet Archive
- Space Apps Challenge
- Project Repository
- uv Documentation
- Moonrepo Documentation
Made with ❤️ by AstroBytes UTB for NASA Space Apps Challenge 2025