Hello

I am a software engineer and researcher. My work focuses on applying novel techniques in software engineering to create tools for automatically configuring, repairing, and managing applications and application infrastructure.

Research

Infer — Automatic configuration of computing environments

Code configuration is hard. Studies have shown that only about a quarter of code snippets found on websites like Stack Overflow and GitHub's gist system are directly executable. Many of these snippets fail to execute because of missing code dependencies, services, or environment variables. Infer focuses on automatically generating computing environments and configuration scripts for arbitrary code bases.

Gistable

An evaluation of the executability of Python code snippets with a focus on dependency errors. ICSME 2018.
ArXiv Preprint IEEE Xplore Source Code

DockerizeMe

Automatically infer a code snippet's direct and transitive dependencies. Also provide an installation ordering that respects transitive dependence at compile time. ICSE 2019.
ArXiv Preprint IEEE Xplore Source Code

V2

Detect instances of configuration drift in code snippets by identifying failures caused by bad dependency versions. Find dependency versions that enable execution. ASE 2019.
ArXiv Preprint IEEE Xplore Source Code

Dozer

Automatically migrate shell commands into high-level configuration languages like Ansible. ICSE 2022.
ArXiv Preprint Source Code

Comparing algorithms for Odd Cycle Transversal — Practical OCT

Graph bipartization (Odd Cycle Transversal) is a well studied problem with applications in embedding problems for near-term quantum computing hardware. Practical OCT applies existing preprocessing techniques for odd cycle transversal and vertex cover, comparing combinatorial branching algorithms against integer linear programming solvers.

ArXiv Preprint ACM Source Code

Spiderdonuts — Analyzing deceptive centrality measures

Walk entropy measures how structured or uniform a graph is in terms of its walk structure, the characteristics of closed walks which can be performed from each vertex. Spiderdonuts studies the relationship of a graph's walk structure to its matrix-based centrality measures, showing that some "deceptive" graphs have different walk classes assigned the same value by well studied centrality measures. Published in LAA.

ArXiv Preprint ScienceDirect Source Code

Investigating effect of Selenium configuration on test flakiness — Wait Wait

Selenium WebDriver is a popular library for performing automated UI testing of web-based applications using continuous integration. However, Selenium tests are often flaky, failing because a test condition is evaluated before the browser has finished rendering the page. Test flakiness can be addressed by giving Selenium a wait condition. Wait Wait compares different Selenium configurations, including choice of waiting strategy, for test stability and runtime performance. AST 2019.

ACM Source Code

Find me on...