Maxim Berman
Research Engineer at Mistral
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolMaxim Berman is a research software engineer based in Zurich with 11 years of experience bridging academic rigor and production-grade ML engineering at Google, Apple, Amazon and Facebook AI. Trained in mathematics, physics and engineering (ENS, UPMC, KU Leuven), he specializes in applied research for computer vision and segmentation, turning state-of-the-art losses and models into reliable PyTorch and TensorFlow implementations. His open-source contribution to the widely cited Lovász-Softmax loss demonstrates both deep theoretical understanding and practical attention to code quality, compatibility, and demos for binary and multiclass segmentation. Comfortable moving between research prototypes and deployable systems, he brings a track record of shipping reproducible tooling and refactoring legacy code for modern frameworks. Colleagues value his blend of academic depth and pragmatic engineering, and his background in statistical physics often surfaces in principled approaches to model design and optimization.
11 years of coding experience
8 years of employment as a software developer
Master of Science (MS), Machine Learning, Vision, Mathematics, Applied Mathematics, Master of Science (MS), Machine Learning, Vision, Mathematics, Applied Mathematics at École Normale Supérieure de CachanBachelor of Science (BSc), Mathematics, Bachelor of Science (BSc), Mathematics at Pierre and Marie Curie UniversityPhD, Engineering Science, PhD, Engineering Science at KU LeuvenClasses préparatoires MP*, Maths, Physics, Classes préparatoires MP*, Maths, Physics at Lycée Louis le Grand, ParisMaster of Science (MSc), Quantum Physics, Master of Science (MSc), Quantum Physics at Ecole normale supérieure
Dutch, English, French, German
Stackoverflow
7,407reputation
1.6mreached
21answers
2questions
command-line-arguments
top-1%
argparse
top-1%
python
top-1%
Github Skills (16)
argparse10
neural-network10
command-line-arguments10
pytorch10
image-segmentation10
loss-functions10
segmentation10
tensorflow10
python10
jupyter-notebook9
lua6
matlab6
paste6
tiff6
rectangles6
Programming languages (14)
Github contributions (5)
bermanmaxim/LovaszSoftmax
Feb 2018 - Feb 2019
Code for the Lovász-Softmax loss (CVPR 2018)
Role in this project:
ML Engineer
ML EngineerContributions:25 commits, 1 PR, 22 pushes in 1 year
Contributions summary:Maxim contributed PyTorch and TensorFlow implementations for the Lovász-Softmax loss function, which aligns with the repository's focus on image segmentation and loss functions. Their work included adding Pytorch loss layers and demos, as well as a TensorFlow version for multiclass segmentation, providing examples for binary and multiclass segmentation problems. The user reorganized code for a better structure and support for newer pytorch/python versions.
softmaximage-segmentationpytorchneural-networksloss-functions
facebookresearch/multigrain
Mar 2019 - Nov 2019
Code for "MultiGrain: a unified image embedding for classes and instances"
Contributions:16 commits, 15 pushes, 8 comments in 7 months
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.