Al Barrentine

About Al Barrentine

Interested in math, systems, and society.Much of my work and research is around text analysis, topic models, location/geo, and natural language processing using real-world public data sets like Wikipedia, Twitter, and OpenStreetMap. Overall, I\'m more interested in building systems that have real impact on people\'s lives than in pure research, though there is a fair amount of original research involved in many of my projects.In the open source world, I\'m the author of the popular international street address parser libpostal. It\'s written primarily in C for portability, uses sequence models to parse messy addresses in dozens of languages all over the world, and trains on over a billion examples from OpenStreetMap and OpenAddresses. The entire pipeline to generate the training data is open source, and the training data have been released as well (one of the largest labeled sequence training sets available to researchers). That project can be found on Github: https://github.com/openvenues/libpostal.Other areas that I work with on a regular basis are collaborative filtering, matrix/tensor factorization, recommenders, network analysis, information retrieval, language modeling, traditional supervised classification/regression and various learning to rank algorithms. I tend to use semi-supervised learning approac for some of the partially or incorrectly labeled datasets that come up in practice.Lastly, I usually build learning systems rather than standalone analyses or one-off models that improve standard benchmarks. Sometimes this means implementations that encourage sparsity so that a model can run entirely in-memory. Often it means utilizing my other true loves: database design and distributed systems, for storing/updating sufficient statistics and online retrieval to serve predictions.

Experience

1. Director of Technology & Data Science

   Florida Rights Restoration Coalition

   Nov 2018 — Present

   Applying new technologies and data science to empower returning citizens (people with past felony convictions) in Florida with the FRRC, an organization led by directly impacted people that was the core group behind passing Amendment 4 to restore voting rights to over 1.4 million people with a past conviction.

Education

• NYU Stern School of Business

  B.S., Social Entrepreneurship, Management & Organizational Behavior, International Business

  2004 — 2008

Skills

• Set Theory
• Algorithms
• Graphical Models
• Numpy
• Unsupervised Learning
• Databases
• Predictive Analytics
• Scipy
• Dimensionality Reduction
• Mysql
• Distributed Systems
• Natural Language Processing
• Recommender Systems
• Mapreduce
• Semi-Supervised Learning
• Information Retrieval
• Topic Models
• C
• Machine Learning
• Conceptual Modeling
• Bayesian Inference
• Distributed File Systems
• Graph Theory
• Collaborative Filtering
• Python