I leverage my background in computer science/data science to help improve the methods by which public policy research is done at my company.I manage MDRC’s cloud-based experimental lab entitled the “DataLab”. Managing involves communicating with users (i.e. the researchers) about their needs and wants, keeping software and associated packages up-to-date (e.g. using packrat and checkpoint to keep R package updates smooth and non-disruptive), and conducting educational sessions with researchers about new data analysis techniques and new software. To that end, I have conducted several educational sessions (e.g. using the tidyverse in R, doing geospatial analysis in R and Python, what is metaprogramming/non-standard evaluation, techniques for doing textual analysis: topic modelling and word embeddings).In addition to managing the DataLab, I am working on a project to create a data portal linking together publicly available data from various government agencies. Steps include: data collection and documentation using R and Python, database configuration using SQL Server, server configuration using either Django or Flask, and front-end configuration/visualization using Javascript + D3 + Leaflet.I'm also currently assisting a research project conduct a causal investigation of the Supervised Release alternative to bail program in NYC. Assistance involves running a multivariate regression model, finding associated statistical significance, checking for robustness, and visualizing results.Along with my responsibilities mentioned above, I also serve on my organization's IRB (Institutional Review Board) which is tasked with ensuring that all MDRC research is in compliance with federal research ethical guidelines. In this role, I review all research which is flagged as needing IRB approval, I read the documentation provided by the team about the nature of the research project, and I consult the IRB federal guidelines to ensure all research abides by said guidelines.

Worked in the Computational Social Science division within the Microsoft Research lab which uses quantitative experimental and non-experimental methods to understand human interactions.Designed a survey to learn about peoples’ interactions within their social networks. The survey was designed through SurveyGizmo, a pilot version was ran on Amazon’s Mechanical Turk, and the full version was launched on YouGov. Subsequent data visualization was done in R using ggplot2, and subsequent data analysis was done in R using multivariate regression with weights obtained by "raking" the survey data using the "survey" package in R.Worked on developing a machine learning model to predict classification categories for historical cables sent to and from the US State Department. The prediction task was trying to predict whether or not a cable was selected for inclusion in a historical archive many years later. Logistic regression, random forests, and boosted decision trees were explored as potential machine learning models using the R programming language. Python was used primarily for data cleaning. Additionally, a Latent Dirichlet Allocation topic model was created using the corpus of cables in the R programming language to expand the feature space.

Worked with researchers in the Center for Decision Sciences within the Business School who were interested in figuring out what tasks were well-suited for "crowdsourcing".Built and ran daily/weekly web scraping tools using beautifulsoup and Selenium in Python and using RESTful APIs to collect data from Kickstarter and Thingiverse (two websites where communities of people come together to solve problems collaboratively).Managed the collection of data from aforementioned web scraping tools in a MySQL database.Aided in the running of the analysis of the aforementioned data using Amazon Web Services to partition new machines to speed up the processing time of the analysis.