Projects

DevOps for Azure API Management

Currently, the management experience for API Management in Azure is highly manual. For teams that want to automate their development and deployment processes, this is not acceptable. Along with a small group inside of Microsoft, I am developing a GitHub Action which will automatically update API Mangement when an backend Azure Function is created or changed. This should help improve the DevOps process for the combination of API Management and Azure Functions.

💻 Code and Roadmap available on GitHub

React + OAuth2

This repository contains a sample React application which implements an OAuth2 PKCE flow with Azure Active Directory as an identity provider. The demo is based on the code from the MSAL sample application, but the PKCE flow is implemented from scratch without assistance from any external libraries.

💻 Code available on GitHub

Feature-Relevant Data Reduction

Over a summer spent at Oak Ridge National Lab and the following fall, I collaborated with researchers to explore data reduction (compression) and its applications in SuperComputing. Specifically, I ran tests and wrote a paper (below) on how lossy compression algorithms could be applied to large datasets without significant impact on the actual features of interest in the data. The concept of dynamically changing compression parameters over the course of a simulation proved promising, though I have not seen it implemented in practice. Fair warning, if you attempt to read the actual paper, my favorite comment from the review process was along the lines: "It would be an interesting result if the writing were not so atrocious." I worked in subsequent drafts to fix this, though I admit the paper is a bit dense. I presented the results at SuperComputing 2018 in Dallas.

🧠 Paper available here

Confidence Scanner

The Confidence Scanner was a project completed in collaboration with the Cognitive and Neural Dynamics Lab at UC San Diego. We built-on existing Natural Language Processing techniques to measure "confidence" in scientific liturature, both primary (journal articles) and secondary (press releases). Ultimately, our key finding was that the further you get from primary sources, there is a tendancy to overstate confidence in scientific results. The project consists of a Python code base built to collect (via scraping and public APIs) and analyze thousands of primary and secondary scientific articles. There is also a poster and paper presented at CogSci 2018 in Madison, Wisconsin.

💻 Code available on GitHub

🧠 Paper available on the conference website

📌 Poster available here