Stef Lindall
Principal Engineer at Modular
Berkeley, California, United States
Summary
🤩
Rockstar🎓
Top SchoolStef Lindall is an experienced software and machine learning architect with 14 years of building production-scale ML systems and inference platforms, currently leading architecture for the MAX accelerator programming platform in Berkeley. He has deep hands-on history at Facebook/Meta and Stripe where he led PyText training, deployed large multilingual transformer models, invented transformer-based representation learning for mixed tabular and sequential data (patented), and shipped high-accuracy fraud and merchant-detection models. Stef combines low-level systems work—autoscaling and sharding inference infrastructure that ran millions of replicas—with data-pipeline and tooling contributions (including improvements to the widely used PyText framework and TorchScript export). Equally comfortable in research and shipping at scale, he has repeatedly turned experimental model ideas into resilient, high-throughput production services.
15 years of coding experience
12 years of employment as a software developer
Bachelor of Science (BS) Mathematics and Computer Science, Bachelor of Science (BS) Mathematics and Computer Science at University of Minnesota
Github Skills (10)
pytorch10
data-pipelines10
data-pipeline10
python10
natural-language-processing9
documentation9
nlp9
torchscript9
machine-learning8
unit-testing7
Programming languages (5)
Github contributions (5)
facebookresearch/pytext
Dec 2018 - Oct 2019
A natural language modeling framework based on PyTorch
Role in this project:
Back-end Developer
Back-end DeveloperContributions:39 commits, 51 PRs, 9 comments in 10 months
Contributions summary:Stef focused on improving the documentation generation process for the pytext framework by modifying the configuration files to host docs on ReadTheDocs. They also refactored the code by moving the random seed configuration into the main configuration file. Furthermore, the user contributed to the data handler rewrite by implementing data source and tensorizer utilities, which are core components of the data pipeline. They also addressed a unit test failure and implemented torchscript export functionality.
language-modelingpytorch
bethebunny/conflag
May 2015 - Apr 2023
Contributions:30 pushes, 1 branch in 8 years
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.