Build

Projects

Coursework, research, and things I built to understand how something actually works — classifiers, causal estimators, a version control system written from scratch. Source for all of it is on GitHub.

Machine Learning & NLP

Supervised models and text analysis, end to end — features, training, evaluation.

  • SpamEmailClassifier

    Spam detection model trained with scikit-learn and evaluated with ROC curves.

    • scikit-learn
    • Classification
    • ROC Analysis
  • Twitter-API-Text-Analysis

    Sentiment analysis of public figures’ tweets using VADER scoring.

    • NLP
    • VADER
    • Sentiment Analysis

Statistics & Causal Inference

Where the interesting part is the assumption, not the model.

  • CausalInferenceJobTraining

    Unconfoundedness vs. independence assumptions, applied to job training and income outcomes.

    • Causal Inference
    • Unconfoundedness
    • Policy Evaluation
  • GLMBootstrappingHurricane

    Generalized linear models and bootstrapping methods applied to hurricane data.

    • GLM
    • Bootstrapping
    • Statistical Modeling

Systems & Data Infrastructure

Rebuilding the tools underneath, to see what they’re made of.

  • Gitlet

    A miniature version control system implementing Git’s core commands with SHA-1 content addressing.

    • Version Control
    • SHA-1
    • Data Structures
  • MongoDBYelpReviews

    Yelp review dataset analysis in MongoDB, benchmarked against an equivalent Pandas approach.

    • MongoDB
    • NoSQL
    • Pandas