I’m Felix Hang, and this blog is a summary of my work in the Fusion 10 Cohort. Last year, I developed Astro AI — an AI application for exoplanet habitability analysis.

The Primary goal of the project was to give definitive, instant habitability scores/answers from sparse astronomical data and make analysis accessible via natural-language queries (e.g., “How habitable is TRAPPIST-1 e?”). In the end, I successfully developed a conversational chatbot interface equipped with AI-powered analysis of NASA Exoplanet Archive data and visualizations to convey habitability factors.

Visualizations included:

• Temperature range (equilibrium temperature context)
• Similarity graph (compare planets to Earth)
• Orbital path vs. habitable zone
• Planet explanation (key habitability characteristics)
• Major challenges:

Implementation:

• Time constraints prevented custom model training.
• Language model occasionally ignored format instructions and produced inconsistent output.
• Visualizations required most development time; initial text-only outputs were hard for users to interpret.
• Handling ambiguous or very new exoplanet queries.

Testing & feedback:

• User testing (including at University of Victoria) emphasized the need for clear visuals and better uncertainty handling.

If you’re interested in the full report, you can access my full project description here.