E xo-Explorer, an AI-powered web app that cleans, engineers, and analyzes NASA's extensive exo-planetary data to predict habitability scores. Built with GPT-4o, LangGraph, and H2O AutoML, it automates scientific discovery, combining rule-based filtering with probabilistic modeling to surface potentially life-supporting worlds across the galaxy.
🧠 Local AI Architecture
- Streamlit: Frontend interface for interactive exploration
- LangGraph + GPT-4o: Orchestrates a modular data-cleaning agent with memory and human-in-the-loop review
- Pandas + SQLite: Stores planetary data and model outputs locally
- H2O AutoML: Trains models to predict exoplanet habitability with automated feature selection and tuning
- OpenAI + LangChain: Powers agent-based code generation and reasoning
- Local CSV & SQL Storage: Persist cleaned datasets and predictions for download or analysis
Backend
Python
SQL Alchemy
Frontend
Streamlit
AI/ML
GPT-4o
AI Data Science Team
LangChain
LangGraph
H2O AutoML
CI / CD
GitHub Actions
Let’s Connect
If you’re a builder, dreamer, or data explorer, reach out. Whether it's a conversation, a collaboration, or just curiosity, I’d love to connect. You can find me on LinkedIn or email me directly.
🧪 How It Works
Step 1: Open the App
ExoExplorer uses NASA Exoplanet Archive
Users first visualize the dataset in a Streamlit dashboard. The app connects to the NASA Exoplanet Archive and it ready for H2O AutoML predictive analysis.
Step 2: Click on Train Model
H2O AutoML Begins Training
The user then clicks on Start Training, which triggers the H2O AutoML pipeline. This pipeline automatically cleans, engineers, and trains a model on the dataset using LangGraph and GPT-4o.
Step 3: AutoML Leaderboard generated
A Best Model is Chosen
The pipeline runs a series of data cleaning and feature engineering steps, generating a leaderboard of models. The best-performing model is selected based on metrics like accuracy and F1 score.
Step 4: Probability-Habitability Column now Fulfilled
Each Planet Record now Contains Habitability Score/h4>
The best model's predictions are saved back to the SQLite database, enriching each planet record with a habitability score. This allows users to filter and explore potentially habitable worlds.
Step 5: Explore The Exoplanet
Planets With Highest Habitability Score
Explore the actual worlds with the highest habitability scores. Visit the NASA Exoplanet Archive to view the full details of each planet, including its characteristics and potenital for supporting life.
📁 Repository Structure
ExoExplorer/
│
├── Exo_Explorer.py # Streamlit app for exoplanet habitability
│
├── cleaning_agent/ # LangGraph-powered data cleaning agent
│ ├── agent.py # Main agent logic and pipeline
│ ├── utils.py # Helper functions
│
├── dataset/ # NASA exoplanet CSV dataset
├── database/ # SQLite database for storing results
├── models/ # Saved H2O AutoML models
├── img/ # Images or diagrams (optional)
│
├── templates/ # Prompt templates for agent instructions
├── parsers/ # Output parsers for GPT responses
├── tools/ # Shared tools/utilities
├── utils/ # General-purpose utility scripts
│
├── README.md # Project overview and instructions
├── requirements.txt # Core Python dependencies
├── additional-requirements.txt # Extra deps for experiments or modules
└── .gitignore # Ignore Python caches, models, etc.
📁 Run Locally
git clone https://github.com/LannonTheCannon/gen_ai_bootcamp.git cd Week3 python3 -m venv venv source venv/bin/activate pip install -r requirements.txt streamlit run Exo_Explorer.py