I'm a data scientist with a management consulting background and a passion for using machine learning, AI, and data to do good in the world.
I'm more or less always hacking away at something and trying to pick up news skills to add to my arsenal. What follows is a selection of some of my experiments (some still in progress, others mostly dead) that I thought people might find interesting.
Photograph: DeepMind
My personal web dashboard / new tab screen. Amongst other features, it automatically checks for and pulls any new stories from Reddit, YouTube, ESPN, and Hacker News. Reddit posts are additionally ranked by a model trained on my historical data.
Added Amazon Alexa like voice functionality to my lil ai that can be used to performing basic web browsing actions like Google searching, navigating between pages, scrolling, zooming, etc.
My lil ai search assistant personalizes my Google search results, bumping up the links I'm most likely to click. For common searches, it can even navigate straight to the desired page skipping the 10 blue links entirely.
As a data professional and hobbyist, I'm often saddened by what little access companies give us to our own data. Given this, I take steps to track and / or acquire it myself.
I appreciate an elegant, well designed API that is clear and just works. Unfortunately, this is rarely the case with AWS products, so I designed a DynamoDB Helper that allows you to interact with a DynamoDB table as simply and easily as a Python dictionary.
AWS Lambda and serverless computing has amazing potential in the data, machine learning, and AI space, however AWS likes to overcomplicate things (see also my DynamoDB Helper) so I made a Lambda Helper that makes deploying serverless functions to AWS easier.
Note: This project currently heavily leverages Lambda functions that are deployed on my personal AWS account. So for now, I'm not making this repo public.
Wanting to learn more about Monte Carlo Tree Search (MCTS), a major part of the original AlphaGo algorithm / paper, I ported a Python 2 implementation I found online over to Python 3. I also created my own implementation to solve some of the drawbacks I observed in the original. For both of these cases, I used Tic-Tac-Toe as my simple, test-case game.
Natural language processing (NLP) / natural language understanding (NLU) is a crucial area of AI research. Once computers can truly understand natural language with all its subtleties and nuance, we will basically have artificial general intelligence (AGI).
Healthcare stands out as the industry where AI can have the most profoundly positive effect on humanity: enabling more accurate, informed health decisions, democratizing access to the best doctors / medical knowledge, and accelerating medical research.
Currently the most promising approach for building strong AI seems to be combining classical AI techniques with machine learning (for example, Google DeepMind's AlphaGo with it's combination of classical tree search and deep learning).
I believe we each have a right to the data we generate. However, companies are often don't give us access to this data or otherwise make it difficult obtain. As such, I enjoy building tools and processes to "acquire" my rightfully owned data.
Economics has much to offer the world of AI. In particular, I'm interested in things like building complex, human-like utility functions, advanced game theory (repeated games, multi-agent problems, mixed-strategy, simultaneous movement, etc.), and AI economies (economies where all the actors are AIs)
Strategic Analyst II (Internal Data Science Consultant)
New York, NY • 2016 - Present
Previously: Strategic Analyst; Freelance Data Scientist
Artificial Intelligence Nanodegree
Online • May - Nov. 2017
Project based, 6 month online program covering “classical” AI, deep learning, and their applications
Data Scientist
New York, NY • Fall 2015
Intensive, project based 12 week data science bootcamp
Associate Consultant
Chicago, IL • 2012 - 2015
Previously: Business Associate
BBA, Economics & Marketing
Notre Dame, IN • 2008 - 2012
Magna Cum Laude (GPA: 3.8/4.0)