🚀 Project Overview: Exo-Traveler

The Fascination & The Itch

My fascination with space began not with data points, but with a question: What would it actually feel like to stand on another world? For years, I've had this persistent "itch"—a compelling urge to work on exoplanets that goes beyond abstract analysis. We've discovered over 6,000 confirmed planets, yet they remain trapped as sterile rows in NASA databases. I wanted to feel them, to imagine the weight of their gravity, the color of their skies, the length of their days. This project is my answer to that longing.

The Problem

Exoplanet data is scientifically rich but experientially hollow. A casual space enthusiast sees pl_eqt: 473K and learns nothing. A sci-fi fan reads about Tatooine's binary suns and has no idea such systems exist in reality. The gap between hard science and human imagination remains unbridged.

My Approach

Exo-Traveler functions as an Interstellar Travel Agency, translating NASA telemetry into immersive, human-centric experiences through three core innovations:

1. Sci-Fi Archetype Matching Engine: Users don't filter by "Temperature > 300K"—they ask to visit Mustafar, Tatooine, or Pandora. A hybrid recommendation algorithm (Cosine Similarity + Euclidean Distance with weighted Z-scores) finds the closest real-world scientific matches to fictional worlds.

2. Multi-Dimensional Filtering Pipeline: Combines strict astrophysical constraints (distance, orbital stability, radiation safety) with iterative fuzzy logic that progressively relaxes tolerances (10% → 40%) when exact matches don't exist, ensuring the system always finds something worth exploring.

3. Immersive Narrative Generation: Integrates Mistral AI to transform raw data into cinematic mission briefings, complete with threat assessments, relativistic time dilation calculations, and sensory descriptions that make abstract numbers visceral.

Methodology

Data Foundation: NASA Exoplanet Archive (6,000+ confirmed planets). Cleaned missing physics using empirical power-law mass-radius relationships, reconstructing 1,800+ otherwise unusable entries.

Ranking Algorithm: Hybrid scoring balances "vibe" (aesthetic/experiential match via normalized feature vectors) with "safety" (habitability metrics). Applies data integrity penalties for incomplete records.

Physics Engine: Calculates Earth Similarity Index (ESI), surface gravity (g = GM/R²), planet type classification (Rocky/Gas Giant/Ocean World), and relativistic travel time using Lorentz transformations.

Key Findings

1. Data Archaeology: ~30% of NASA's 6,000+ exoplanets lacked mass/radius data. Power-law reconstruction (M ∝ R^3.7) salvaged 1,800+ entries, tripling the viable pool. Data integrity penalties prevent reconstructed planets from dominating rankings.

2. Astrophysical Patterns: Temperature follows Stefan-Boltzmann law (T⁴ ∝ L) with R² > 0.85. Rocky planets cluster in the 0.2–10g gravity range (human-tolerable); Gas Giants span 7 orders of magnitude. The habitable zone (0–50°C) represents <5% of discovered planets.

3. Algorithmic Performance: Hybrid ranking (60% vibe via cosine similarity, 40% safety via Euclidean distance) achieved 92%+ confidence for clear archetypes like Mustafar and Mann's Planet. Iterative fuzzy logic rescued 87% of searches that would've failed under strict filtering.

4. Experiential: Real analogs exist for major sci-fi worlds. Proxima Centauri d = Pandora (4.24 ly, tidally locked red dwarf). LHS 1140 b = Miller's Planet (JWST-confirmed ocean world, 1.9g). HD 20781 = Tatooine (binary sunset). K2-233 b = Mustafar (lava hellscape).

5. Relativistic Costs: At 0.82c, Proxima Centauri d takes 5.1 Earth years but only 2.9 ship years (43% time dilation). For 500+ ly targets, crew ages decades while Earth centuries pass—an existential asymmetry the project quantifies viscerally.

6. Design Impact: "Threat Monitors" gamify danger (3.2g planet → 85% bar labeled "GRAVITY CRUSH"). Match confidence percentages and uniqueness scores ensure transparency: users always know how certain the system is, not just what it recommends.

Bottom Line: This isn't a dashboard. It's a portal. Every filter adjustment, every chart, every AI-generated transmission asks the same question that's driven me from the start: What would it feel like to be there?

Built With

• csv
• hex
• hexexplorecharts
• math
• mistral-ai-api
• nasa-exoplanet-archive
• numpy
• pandas
• python
• scikit-learn
• snowflake
• sql