← Back to Portfolio
DataOrbit ML Model
Winner of the Predictive Modeling Award at UCSB's DataOrbit 2025 hackathon. Built and evaluated a machine learning model to predict outcomes from raw data, using feature engineering and cross-validation to maximize accuracy.
The Problem
A hackathon dataset arrived raw and unmodeled, with a fixed deadline and no guarantee that any signal in it was easy to reach.
Approach
Prioritized feature engineering over model complexity — cleaned and reshaped the raw data with pandas and NumPy, then trained and compared scikit-learn models using cross-validation so the reported accuracy reflected generalization rather than a lucky split.
The Outcome
Won the Predictive Modeling Award at DataOrbit 2025.
Highlights
- ▹Winner, Predictive Modeling Award at DataOrbit 2025
- ▹Feature engineering pipeline built with pandas and NumPy
- ▹Cross-validated model selection to avoid overfitting to a single split
Stack
Python
scikit-learn
pandas
NumPy