Justus Perillieux
Staff Research Engineer at Lightning AI
Dusseldorf, North Rhine-Westphalia, Germany
Summary
🤩
Rockstar🎓
Top SchoolJustus Perillieux is a Staff Research Engineer based in Düsseldorf with a decade of experience building scalable deep learning systems and production-ready tooling for PyTorch ecosystems. He has driven metric, logging, and DDP improvements across high-profile open-source projects including PyTorch Lightning, torchmetrics, and Ignite, focusing on maintainability, type-safety, and distributed correctness. His background spans medical imaging research and industrial work, bringing practical domain expertise in deploying ML for healthcare alongside large-model scaling. At Lightning AI he moved from project leadership to staff-level research engineering, continuing to bridge research and engineering to make complex models reliable in multi-GPU and CI/CD environments. Colleagues rely on him for thoughtful refactors and tests that reduce tech debt while enabling reproducible, distributed training.
10 years of coding experience
6 years of employment as a software developer
Master of Science - MS, Electrical Engineering, Information Technology and Computer Engineering, Master of Science - MS, Electrical Engineering, Information Technology and Computer Engineering at RWTH Aachen University
German, English, French
Github Skills (26)
pytorch10
lib10
python10
data-science10
image-classification10
testing10
machine-learning10
pytorch-lightning10
image-segmentation10
deeplearning-ai10
deep-learning10
segmentation10
neural-network10
computer-vision10
refactoring10
Programming languages (13)
Github contributions (5)
Lightning-AI/pytorch-lightning
Mar 2020 - Jan 2023
Pretrain, finetune ANY AI model of ANY size on multiple GPUs, TPUs with zero code changes.
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 2886 reviews, 111 commits in 2 years 10 months
Contributions summary:Justus's contributions focused on modifying the PyTorch Lightning library to include and improve metric logging. Their commits added support for logging metrics with hyperparameters using TensorBoard, implemented bug fixes for hyperparameter logging with metrics, and expanded metric functionality to include averaging across multiple devices in a DDP setting. The user also introduced and refactored Sklearn and Native Torch metric classes.
pythonheadachespytorch-modelsdata-sciencehandling
Lightning-AI/torchmetrics
May 2020 - Jan 2023
Machine learning metrics for distributed, scalable PyTorch applications.
Role in this project:ML Engineer
ML EngineerContributions:556 reviews, 33 commits, 31 PRs in 2 years 8 months
Contributions summary:Justus's primary contributions revolved around the development and implementation of machine learning metrics for PyTorch applications. They introduced new metric classes, including utilities, tests, and documentation. Their work involved refactoring existing metrics (e.g., for scikit-learn compatibility) and integrating native PyTorch implementations. The user demonstrated a strong focus on ensuring correct tensor conversions, DDP synchronization, and metric aggregation.
pytorchscalablepythonanalysesdata-science
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.