Intuix

Exoplanet surveys generate large datasets that need careful classification and interpretable results.

The project combines data from Kepler, TESS, and K2, using physics-informed feature engineering, a LightGBM classifier, and SHAP explanations. SMOTE is applied within training folds to address class imbalance without contaminating validation data.

• A FastAPI inference interface and Streamlit exploration UI.
• Explanations of model predictions with SHAP.
• NASA Space Apps Challenge 2025 Global Finalist Nominee, as listed in my latest resume.