Max Gerlach
Manager Research Engineering - High Performance Computing at DeepL
Cologne, North Rhine-Westphalia, Germany
Summary
🤩
Rockstar🎓
Top SchoolMax Gerlach is a manager of research engineering specializing in high-performance computing with 13 years of experience bridging academic research and production ML infrastructure. Based in Cologne, he progressed from a PhD in theoretical physics to leading HPC efforts at DeepL, where he now oversees scalable training systems and research-to-production workflows. His hands-on background includes improving distributed training tests and fixes in the popular Horovod project, revealing subtle NCCL-related edge cases and stabilizing PyTorch/MXNet test suites. Prior roles at Universität zu Köln and a visiting stint at the Weizmann Institute reflect deep theoretical foundations combined with practical engineering. Colleagues describe him as someone who converts complex numerical problems into robust, testable systems and who values improving test coverage to surface hard-to-find production bugs.
13 years of coding experience
11 years of employment as a software developer
Dr. rer. nat. (Ph.D.) Theoretical Physics, Dr. rer. nat. (Ph.D.) Theoretical Physics at University of CologneUniversity of MilanDiplom (M.Sc.) Physics, Diplom (M.Sc.) Physics at Leipzig UniversityItalian Language, Italian Language at Università per Stranieri di Perugia
English, German, Italian, French
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (11)
mxnet10
pytorch10
distributed-training10
tensorflow10
horovod10
test-automation10
testing10
mpi9
ccl8
machine-learning7
deep-learning7
Programming languages (15)
Github contributions (5)
horovod/horovod
Sep 2020 - Jan 2023
Distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.
Role in this project:
Back-end & Test Automation Engineer
Back-end & Test Automation EngineerContributions:3 releases, 245 reviews, 68 commits in 2 years 5 months
Contributions summary:Max primarily contributed to the testing framework of the Horovod project. Their work included fixing environment variable usage in alltoall tests and adding new tests for alltoall edge cases, which uncovered issues with specific NCCL versions. The user also cleaned up MXNet broadcast tests by renaming functions and correcting logging. Furthermore, the user made improvements to PyTorch tests by removing stalls and adding the return of join calls, along with implementing fixes to reducescatter and grouped_reducescatter operations.
mpikeras-tensorflowtrainingbaidutensorflow
maxhgerlach/slurm-scripts
May 2013 - Feb 2017
Contributions:42 commits, 5 pushes in 3 years 9 months
pythonbashsimple-shellclustershell
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.