I’m a mathematician seeking employment as a data scientist. I’m very versatile in my ability to apply what I know: I understand ideas at a high level of generality, which enables me to recognize non-obvious ways of applying them to powerful effect.
My skills
Machine learning: I’ve made extensive use of linear models for regression and classification, Bayesian hierarchical models, principal component analysis and cross-validation. I’ve also experimented with random forests and collaborative filtering, and am familiar with their strengths and limitations.
General analytics: I worked as a research analyst at GiveWell vetting cost-effectiveness estimates of global health interventions (example). I’ve done economic analyses in various capacities, including a thorough review of the evidence for and against majoring in economics increasing earnings (link) that I wrote at Cognito Mentoring to advise high school students.
Programming: In addition to using R for data science, I’ve done full stack web development in Ruby on Rails, SQL, Javascript and Backbone.js. I can quickly become fluent in any programming language. I have code samples on Github and describe my largest projects below.
Communication: I have a strong interest in communicating to technical ideas to less technical audiences. I’ve been deeply involved in math education in many capacities: high school and college instruction, online courses, personal tutoring and expository writing.
Projects
Speed Dating Project (Source ): I analyzed a public speed dating data set to predict participants’ decisions. I gradually shifted focus to understanding what the data tells us about human diversity. Following this line of thought ultimately facilitated the creation of a better predictive model, while having relevance extending beyond the context of speed dating.
• Constructed a Bayesian hierarchical modeling of individual preferences.
• Derived predictive features using revealed preferences, collaborative filtering, and PCA.
• Discovered and quantified substantial and statistical robust variation between individuals with respect to attractiveness / personality trait tradeoffs.
• Used PCA to identify demographic clusters with unusually high preferences for attractiveness and for intelligence.
• Determined the extent to which the group consensus on somebody’s attractiveness predicted individuals’ perceptions.
StackOverflowClone (Source): A clone of the StackExchange network of websites. Single page app written in Backbone.js and Ruby on Rails. Implemented features include:
• Authentication
• Questions, Answers & comments
• Voting & Tagging
• Filtering questions by vote count, author, voter, and tag.
• Editing objects in place
• Auto-complete search
Academic Publications:
Ramanujan congruences for a class of eta quotients (2010)
I investigated a problem adjacent to a famous discovery by Srinivasa Ramanujan: “If n is 4 more than a multiple of 5, the number of partitions of n is divisible by 5.” I proved that a class of superficially similar patterns doesn’t exist. The proof uses the theory of modular forms (mod p).