ExoAI is an AI-powered platform developed for the NASA Space Apps Challenge to automate the classification of exoplanet candidates using open-source data from Kepler, K2, and TESS missions.
The system integrates machine learning, a .NET 9 backend, and a Flutter-based interface to deliver accurate, real-time predictions and improve the efficiency of exoplanet discovery.
Developed an ML model trained on NASA’s Kepler, K2, and TESS datasets using XGBoost and Logistic Regression.
The model classifies signals as confirmed planets, candidates, or false positives with an accuracy exceeding 95%.
Built a secure and scalable REST API for predictions, data management, and model retraining.
Implemented with Entity Framework Core, JWT authentication, and Docker for deployment.
Created a responsive web dashboard that enables users to upload datasets, visualize predictions, view confidence scores, and retrain the model interactively.
Automates exoplanet identification, reducing manual analysis time by approximately 80%.
It provides researchers and students with direct access to NASA’s datasets through an accessible, user-friendly interface, accelerating the pace of exoplanet discovery and promoting open scientific collaboration.
Initial splash screen displaying ExoAI’s logo and entry point to the system.
Allows new users to create an account securely before accessing the platform.
Secure login screen for registered users to access their dashboards.
Main dashboard showing real-time Kepler data insights, statistics, and prediction summaries.
Upload interface for users to submit new datasets in CSV/JSON format for prediction.
Displays prediction results with confidence scores and visual data insights.
Includes sections for:
- Home: Overview and statistics
- Upload: Dataset input and model execution
- Result: Prediction outputs and analytics
- Settings: Model configuration (learning rate, dropout rate, etc.)
- Logout: Secure session termination
Python, scikit-learn, XGBoost, .NET 9, Entity Framework Core, Flutter, Docker, NASA Exoplanet Archive
Navigation menu for Home, Upload, Result, Settings, and Logout.
Model configuration options for learning rate, dropout rate, batch size, and training epochs.
Exo-AI/
├── Exo.API/ # .NET 9 Web API
│ ├── Controllers/ # API endpoints
│ ├── Entities/ # Data models
│ ├── DTOs/ # Data transfer objects
│ ├── Data/ # EF Core DbContext, Migrations
│ ├── Services/ # Business logic, ML integration
│ ├── Program.cs
│ ├── appsettings.json
│ └── Dockerfile
├── Exo.ML/ # Machine learning code & models
│ ├── data/ # raw/, processed/
│ ├── preprocessing/ # Data preprocessing scripts
│ ├── training/ # Training scripts
│ │ └── train.py
│ ├── inference/ # Inference scripts
│ │ └── predict.py
│ ├── models/ # Exported models (pkl/json)
│ └── requirements.txt
├── Exo.Flutter/ # Flutter web dashboard
│ ├── lib/ # Flutter source code
│ │ ├── screens/ # UI screens
│ │ ├── widgets/ # Reusable components
│ │ └── services/ # API client
│ ├── pubspec.yaml
│ └── web/
├── Images/ # README screenshots
├── docker-compose.yml # Multi-container setup
├── .env.example # Environment variable template
├── .gitignore # Git ignore rules
├── LICENSE # MIT License
└── README.md # This file
cd Exo.API
dotnet restore
dotnet build
dotnet run --urls "http://localhost:5000"
# or with Docker
docker build -t exo-api .
docker run -e ASPNETCORE_ENVIRONMENT=Production -p 5000:80 exo-apiEnvironment Variables (see .env.example):
ASPNETCORE_ENVIRONMENT=Development
ConnectionStrings__DefaultConnection=Server=.;Database=ExoAiDb;Trusted_Connection=True;
Jwt__Key=your_super_secret_key_here
cd Exo.ML
python -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windows
pip install -r requirements.txt
python training/train.py # Trains and saves model to models/train.py:
- Loads raw CSV(s) from
data/. - Performs preprocessing and feature engineering.
- Trains XGBoost/LogisticRegression.
- Exports model to
models/withmetadata.json(metrics, hyperparameters).
cd Exo.Flutter
flutter pub get
flutter run -d chromeConfigure API URL in lib/services/api_client.dart.
docker-compose up --build- Register/login via Flutter dashboard.
- Upload dataset (CSV/JSON) → sent to API.
- API processes data via ML module, returns predictions + confidence.
- View results, inspect feature importance, retrain with new hyperparameters.
- Save retrained model for future inference.
- Store large datasets/models externally (e.g., Git LFS).
- Use CI (e.g., GitHub Actions) for tests and validation.
- Version models semantically (e.g., v1.0.0) with
metadata.json. - Secure secrets with environment variables or secret manager.
Developed by a passionate team of tech enthusiasts for the NASA Space Apps Challenge 2025.
MIT License © 2025 ExoAI Team