Data Science Projects

Recidivism Forecasting Challenge (National Institute of Justice, 2021)

https://nij.ojp.gov/funding/recidivism-forecasting-challenge

• Prize winner ($19,500) in a machine learning competition hosted by the National Institute of Justice.
• Predicted recidivism using person- and place-based variables, with the goal of improving outcomes for people serving community supervision sentences.
• Used XGBoost, AdaBoost, LightGBM, CatBoost, autoencoders, and logistic regression in Python.
• The code is available on GitHub.

School Closures and the Digital Learning Market

In this project, we analyzed how COVID-19 school closure orders changed the use of digital learning platforms, using data provided by LearnPlatform. The code and data visualizations are available on Kaggle.

Recommendation System (211 LA County, 2021–2022)

with the 211 LA Data Team

• Developed a recommender system using deep learning and machine learning models in PyTorch and scikit-learn.
• Deployed the model using Flask, Docker, and Google Cloud Platform.
• Developed A/B test plans and managed their implementation.