Mehdi Cherti
Postdoctoral Researcher at Forschungszentrum Jülich
Cologne, North Rhine-Westphalia, Germany
Summary
👤
Senior🎓
Top SchoolMehdi Cherti is a postdoctoral researcher and deep learning specialist with 11 years of experience, holding a PhD in Machine Learning from Université Paris-Saclay. Based at Jülich Supercomputing Center and affiliated with the LAION team, he focuses on large-scale training, generative models, and methods that enable efficient transfer and robust out-of-distribution generalization. His work spans academia and research labs including Helmholtz AI and École des Mines, combining hands-on model engineering with theoretical ML expertise. Notably, he has contributed to open-source efforts around text-to-image transformers—improving training pipelines for a popular DALL·E PyTorch implementation. Comfortable with high-performance compute environments, he brings practical experience in optimizer/scheduler management and reproducible training at scale. Curious and methodical, he often bridges novel research ideas with production-oriented training practices.
11 years of coding experience
4 years of employment as a software developer
Doctor of Philosophy (PhD), Machine learning, Doctor of Philosophy (PhD), Machine learning at Université Paris Sud (Paris XI)Master 2 recherche, Computer science - Machine learning, Master 2 recherche, Computer science - Machine learning at Université René Descartes (Paris V)Engineer's degree, Computer Science, Engineer's degree, Computer Science at INSEABachelor's degree, Mathematics and Computer Science, Bachelor's degree, Mathematics and Computer Science at Faculté des sciences de Rabat
French, English, German
Github Skills (7)
attention-mechanism10
transformers10
pytorch10
artificial-intelligence10
deep-learning10
python10
mlops9
Programming languages (4)
Github contributions (5)
lucidrains/DALLE-pytorch
May 2021 - Jun 2021
Implementation / replication of DALL-E, OpenAI's Text to Image Transformer, in Pytorch
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 8 commits, 3 PRs in 1 month
Contributions summary:Mehdi contributed to the training pipeline of the DALL-E model. Their work involved modifying the `train_dalle.py` script to incorporate features like saving and resuming the optimizer and scheduler states, saving the epoch number, and adding the `attn_types` argument. The user also made minor changes like using quotes instead of double quotes for strings. These changes aimed to improve training management and flexibility.
pytorchmulti-modaltransformersdall-eattention-mechanism
machinedesign/machine_design
Dec 2016 - Aug 2018
Contributions:207 pushes, 1 branch in 1 year 8 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.