Landan Seguin
Senior Research Scientist at Databricks Mosaic Research
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolLandan Seguin is a Senior Research Scientist with eight years of experience building and shipping deep learning research into real-world systems from the San Francisco Bay Area. At Databricks Mosaic Research he has led and executed applied ML work—contributing production-ready features (EfficientNet, DeepLabv3+, weight standardization, stochastic depth) to the Composer training library and running extensive benchmarks that trade off training time and accuracy. His background spans adversarial robustness research at Intel Labs, vision-and-language and VQA modeling at Georgia Tech and MIT Lincoln Lab, and practical deployment work such as optimizing models for Jetson devices and Android. He combines strong experimental rigor (automated testing, reproducible PyTorch codebases, and documented results) with hands-on optimization of training pipelines and dataloaders. An active open-source contributor, he improved DeepLabv3+ performance and dataset tooling in the widely used mosaicml/composer repo. Beyond models, he focuses on interpretability and uncertainty estimation to make research usable and trustworthy in production.
8 years of coding experience
6 years of employment as a software developer
Bachelor's degree Computer Engineering 3.76 / 4.00, Bachelor's degree Computer Engineering 3.76 / 4.00 at Georgia Institute of Technology
Github Skills (13)
neural-network10
computer-vision10
pytorch10
machine-learning10
deep-learning10
python10
model-optimization10
imagenet9
llm9
segmentation9
repr9
image-segmentation9
rep9
Programming languages (2)
Github contributions (5)
mosaicml/composer
Jan 2022 - Oct 2022
Supercharge Your Model Training
Role in this project:
ML Engineer
ML EngineerContributions:193 reviews, 43 commits, 71 PRs in 9 months
Contributions summary:Landan primarily contributed to benchmarking and improving the performance of the DeepLabv3+ model within the `composer` repository. Their work involved implementing and refactoring code for the ADE20k dataset, including the creation of transformations and dataloaders. They also made significant changes to the model training process, optimizing it with techniques such as weight standardization and fixing issues related to loss functions. Key contributions also include updating DeepLabv3+ with the current PyTorch version.
pytorchml-systemsdeep-learningneural-networksmachine-learning
Landanjs/composer
Oct 2021 - Aug 2023
Composing methods for ML training efficiency
Contributions:440 pushes, 129 branches in 1 year 10 months
machine-learningml-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.