Recent Projects

---

Predicting Patent Rejections with NLP

Trained a model to predict rejections under U.S.C. § 101 for claims directed to “abstract ideas.” Claim text was vectorized using word embeddings, and a random forest classifier was used for the final prediction. The final model proved to be 96% accurate on the balanced test set of 32,000 claims, identifying 94.5% of the applications that were rejected.

---

Predicting User Churn

Trained a model for predicting user retention of a ride-share service and provided recommendation for improving user retention.

---

Exploring Accident Data in Denver

Analyzed accident reports over the past 5 years in the city of Denver by looking temporal, geographic, and human factors associated with accident data. Created an interactive map of accident data.

---