Tom Goldstein
Associate Professor
Washington, District of Columbia, United States
Summary
🤩
Rockstar🎓
Top SchoolTom Goldstein is the Perotto Associate Professor of Machine Learning at the University of Maryland with eight years of experience bridging academic research and industry applications in computer vision, NLP, and wireless communications. He has held research roles at Facebook and Zipline, where he built perception and scene-understanding systems for autonomous medical delivery across continents, and maintains active research programs from assistant to associate professor. Tom’s work spans both theoretical applied mathematics (PhD, UCLA) and practical ML systems engineering, including contributions to open-source tools for visualizing neural network loss landscapes that improve portability and parallel computation. Based in Washington, D.C., he combines deep academic rigor with hands-on model and systems development, often tackling the reproducibility and deployment challenges that separate lab results from real-world impact.
8 years of coding experience
3 years of employment as a software developer
Bachelor of Arts (BA), Mathematics, Computer Science, Bachelor of Arts (BA), Mathematics, Computer Science at Washington University in St. LouisMaster of Science (MS), Mathematics, Master of Science (MS), Mathematics at University of California, Los Angeles
Github Skills (10)
mpi10
pytorch10
parallel-computing10
python10
machine-learning9
dataprep8
data-loading8
data-preprocessing8
preprocess8
preprocessing8
Programming languages (3)
Github contributions (5)
tomgoldstein/loss-landscape
Sep 2018 - Jan 2019
Code for visualizing the loss landscape of neural nets
Role in this project:
ML Engineer
ML EngineerContributions:31 commits, 25 pushes, 1 branch in 4 months
Contributions summary:Tom primarily contributed to the project by modifying and refactoring code related to parallel computation and loss landscape visualization. They addressed issues around the dependency on the mpi4pytorch library by introducing a placeholder class for systems without MPI support. Further, the user made changes that involved renaming the 'mpi4pytorch' module to 'mpi', updating the data loading process for random sub-sampling, and fixing typos in comments. Overall, the contributions enhanced the portability and functionality of the loss landscape visualization code.
netsneural-netsvisualizinglosslandscape
tomgoldstein/stone
Aug 2018 - Aug 2018
Contributions:27 commits, 22 pushes in 2 days
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.