Jake Grigsby
Doctoral Student at The University of Texas at Austin
New Canaan, Connecticut, United States
Summary
🤩
Rockstar🎓
Top SchoolJake Grigsby is a PhD student at UT Austin with eight years of experience applying machine learning to real-world and research problems, blending deep learning, reinforcement learning, and time-series forecasting. He has contributed to high-impact open-source projects—improving adversarial NLP tooling in TextAttack and extending Transformer-based spatiotemporal forecasting in Spacetimeformer—demonstrating practical skills in model robustness, data augmentation, and training pipelines. His background includes research roles at UVA QData Lab and IBM as well as hands-on ML work in scientific imaging, reflecting an ability to translate academic ideas into usable code and experiments. Comfortable teaching and mentoring, he also brings strong foundations in mathematics and computer science from UVA, with a knack for adding reproducible utilities like normalization layers and prediction helpers that streamline model development.
8 years of coding experience
3 years of employment as a software developer
Mathematics and Computer Science, Mathematics and Computer Science at University of Virginia
Github Skills (14)
transformers10
pytorch10
eval10
machine-learning10
nlp10
adversarial-attacks10
time-series-forecasting10
deep-learning10
data-augmentation10
python10
evaluation10
model-testing10
data-preprocessing9
wordnet8
Programming languages (2)
Github contributions (5)
QData/spacetimeformer
Sep 2021 - Aug 2022
Multivariate Time Series Forecasting with efficient Transformers. Code for the paper "Long-Range Transformers for Dynamic Spatiotemporal Forecasting."
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 1 review, 63 commits in 11 months
Contributions summary:Jake primarily contributed to the implementation and modification of a time series forecasting model based on Transformer architecture, specifically focusing on the "Spacetimeformer" model. Their work includes the introduction of new functionalities, such as RevIN for input normalization, and the expansion of supported datasets. They also made changes to the model's training loop, including adjustments to learning rate schedulers and loss functions. Furthermore, the user worked on adding utility methods for predictions, including one for handling the output of the models.
pytorchforecastingtransformersefficient-transformersseries-forecasting
QData/TextAttack
Jun 2020 - Jul 2020
TextAttack 🐙 is a Python framework for adversarial attacks, data augmentation, and model training in NLP https://textattack.readthedocs.io/en/master/
Role in this project:ML Engineer
ML EngineerContributions:33 commits, 10 PRs, 28 pushes in 20 days
Contributions summary:Jake focused on implementing and refining adversarial attack techniques within the TextAttack framework. Their commits primarily involved additions and modifications to the `GreedyWordSwapWIR` search method, including the introduction of the "pwws" (Probability Weighted Word Saliency) method. Furthermore, the user added new recipes and transformations related to data augmentation, demonstrating contributions towards improving model robustness and performance through adversarial training and data augmentation. The user also updated documentation.
pythonadversarial-machine-learningadversarialmodel-trainingtraining
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.