India High School Exoplanet Data Challenge
Devpost hackathon: India High School Exoplanet Data Challenge. Imported automatically.
- Year
- 2026
- Winning projects
- 5
- Participants
- 480
- Prize pool
- $10,300
- Platform
- Devpost
Winners
1st Place
- Hunting Exoplanets with Machine Learning1st PlaceA leakage-audited XGBoost that sorts NASA Kepler signals into confirmed planet, candidate, or false positive — 93% F1, fully explainable, and honest about what Python
2nd Place
- Celestivine2nd PlaceExplainable AI for Exoplanet Discovery using Random Forest and SHAP, enabling astronomers to understand why a Kepler Object of Interest is classified as a planePythonStreamlit
3rd Place
- Kepler-AI3rd PlaceCombining raw NASA observational data with advanced physical feature engineering to hunt for exoplanets with 96.8% precision.Python
Audience Choice
- Predicting KOI Disposition Without NASA's Vetting FlagsBest Explainability Award: Submission that makes complex ML most understandable to a layman-audience.You can't put the answer inside the prediction. NASA's vetting flags are the answer, so my leakage-conservative model drops them. Three classes, 0.836 macro F1 JavaScriptPythonReact
Other
- ExoJuryRising Star Award For the strongest solo participant, no team.Every Kepler candidate gets a fair trial: a calibrated verdict, a 95% statistical guarantee, and an AI-written opinion — from a pipeline that found real errors PythonStreamlit