Built for the 2025 NASA Space Apps Challenge — Team TangoisdownHQ
TID-AD-ASTRA is an explainable AI system that analyzes and interprets planetary data from NASA’s Exoplanet Archive and the Open Exoplanet Catalogue.
It predicts potential exoplanet habitability while explaining the why behind every decision — giving scientists, educators, and explorers transparent insight into how AI understands alien worlds.
The name means “To the Stars”, symbolizing our mission to make deep-space data more interpretable, accessible, and open to everyone.
Users can:
- Upload new datasets directly into:
ml/app/data/uploads/
python fetch_data.py --source nasa python fetch_data.py --source open
- Explainable AI for Planetary Habitability
Predicts exoplanet classification and habitability index
Generates interpretable explanations for each prediction
Detects missing or incomplete data and provides contextual diagnostics
- Data Lineage & Provenance
Integrates NASA Exoplanet Archive + Open Exoplanet Catalogue
Tracks dataset origin for each trained model artifact
Ensures transparency in AI learning and decision-making
- Model Management Console (FastAPI)
Central registry with:
Metadata (registry.json)
Lineage tracking
Explainability endpoints (/models/explain, /models/lineage)
Auto-loads the latest trained model and exposes prediction APIs
- CLI Mission Console
Terminal-based “Mission Control” for planetary analysis
Displays:
Prediction summary
Habitability index
Missing-data diagnostics
Dataset provenance
🪐 Example Output
BD+20 2457 b is predicted as class 1 with confidence 0.77. Top influencing factors: feature_1, feature_0. Habitability index: 0.00 — unlikely to support Earth-like life.
Missing or incomplete data for fields: discovery_year, radius_earth, host_star.temperature. Default estimates were used where possible.
Layer Description Data NASA Exoplanet Archive + Open Exoplanet Catalogue Model Scikit-learn classifier with explainability Backend FastAPI service exposing /models, /planets, /explain Interface CLI console + REST API Storage Local model registry (models/artifacts/registry.json)
git clone https://github.com/TangoisdownHQ/TID-AD-ASTRA.git cd TID-AD-ASTRA/ml
python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt
make run
curl -X POST "http://127.0.0.1:8000/models/explain"
-H "Content-Type: application/json"
-d '{"planet_name":"Kepler-442b","features":[]}' | jq
- Launch the Mission Console make run-console
TID-AD-ASTRA supports open exploration — upload your own exoplanet datasets, telescope data, or simulated planetary environments.
- Upload CSV Files
Place your dataset in:
ml/app/data/uploads/ └── your_exoplanet_data.csv
Column Description planet_name Planet identifier mass_earth Mass (in Earth masses) radius_earth Radius (in Earth radii) orbital_period_days Orbital period semi_major_axis_au Semi-major axis equilibrium_temperature_k Temperature (K) eccentricity Orbital eccentricity discovery_year Discovery year distance_from_earth_ly Distance (light-years) host_star_temperature Star temperature (K) host_star_spectral_type Star type (e.g., G2V)
python fetch_data.py --source nasa python fetch_data.py --source open
python -m app.models.classifier --train ml/app/data/uploads/your_exoplanet_data.csv
Source Description NASA Exoplanet Archive Official exoplanet dataset Open Exoplanet Catalogue Community-curated planetary data
Category Technology Language Python 3.11 Framework FastAPI ML Library Scikit-learn Explainability SHAP / Feature Importance Data Layer CSV + JSON registries Interface CLI (Rich-based) + REST API
TID-AD-ASTRA is a foundation for autonomous, explainable AI agents that can:
Reason adaptively about planetary data across star systems
Integrate with future NASA APIs and sensors
Support real-time decision support for interplanetary logistics
🧠 AI & Modeling
Expand models with neural networks and ensemble systems
Add SHAP visual dashboards for interpretability
Enable real-time inference from streaming telemetry
Automate synchronization with NASA’s live API
Merge data from multiple observatories (Kepler, TESS, Gaia)
Visualize data provenance with interactive lineage graphs
Build a web-based Mission Console with dynamic dashboards
Support drag-and-drop dataset uploads
Add 3D planetary system visualizations (using Three.js + NASA JPL data)
Containerize via Docker and deploy on Fly.io or Kubernetes
Add secure API keys and user roles for collaboration
Implement automated retraining and model versioning pipelines
Develop an autonomous science agent capable of reading research papers, updating models, and proposing new exploration targets
Use generative AI to simulate unseen planetary systems
Integrate with spacecraft telemetry for live adaptive AI analytics
“To the stars — and beyond the noise.” 🌠
Role Name / Handle Focus Founder & Engineer @TangoisdownHQ Cybersecurity, AI Infrastructure, Explainability, System Design
💻 GitHub: TangoisdownHQ
🌐 Live Demo: https://tid-adastra.fly.dev
Field Value Region NASA Space Apps Challenge 2025 Team Name TangoisdownHQ Project TID-AD-ASTRA
This project is open-source under the MIT License. NASA datasets and related content are used under the NASA Open Data Policy.