William De Vazelhes
Research Engineer at GenBio AI
Dubai, Dubai, United Arab Emirates
Summary
🤩
Rockstar🎓
Top SchoolWilliam De Vazelhes is a research engineer and PhD candidate at MBZUAI with eight years of hands-on experience in machine learning, optimization, and sparse learning techniques. He has a strong track record bridging research and production: from developing a widely used scikit-learn-compatible metric-learning library (metric-learn, 1.3k+ stars) to contributing documentation and robustness improvements in scikit-learn. His recent work includes optimization and stabilization for large-scale model pretraining (Falcon family) and practical distillation/pruning pipelines at the Technology Innovation Institute. Earlier roles span information-theoretic research at Huawei, CT reconstruction and RL internships, and applied ML systems for fuzzy matching and classification in industry. William combines deep theoretical knowledge (Rate Distortion, optimal transport, VAE) with pragmatic engineering—vectorizing algorithms, fixing numerical issues, and improving memory and speed for real-world use. Based in Dubai, he’s as comfortable tuning large model hyperparameters on cloud clusters as he is proving algorithmic correctness on mathematical toy problems.
8 years of coding experience
4 years of employment as a software developer
Saint Jean de PassyDoctor of Philosophy - PhD, Machine Learning, GPA 3.96/4, Doctor of Philosophy - PhD, Machine Learning, GPA 3.96/4 at MBZUAI (Mohamed bin Zayed University of Artificial Intelligence)Charles III University of Madrid (Universidad Carlos III de Madrid)PCSI-PSI, PCSI-PSI at Collège StanislasMaster of Science - MS, Master of Science - MS at Ecole Supérieure d'Electricité
Spanish, English, French
Stackoverflow
466reputation
49kreached
4answers
0questions
csv
top-5%
Github Skills (19)
pytest10
python10
scikit10
metric-learning10
machine-learning10
numpy10
scikit-learn10
documentation10
csv9
data-structures8
data-structure8
algorithms8
algorithm8
dot-product6
multi-index6
Programming languages (4)
Github contributions (5)
scikit-learn-contrib/metric-learn
May 2018 - Apr 2021
Metric learning algorithms in Python
Role in this project:
ML Engineer
ML EngineerContributions:4 releases, 21 reviews, 74 commits in 3 years
Contributions summary:William primarily contributed to the development and improvement of metric learning algorithms within the `metric-learn` repository. They focused on refactoring and enhancing existing algorithms, such as LMNN, by fixing gradient computations and improving memory performance. Their work also involved adding new features like memory-efficient implementations for algorithms such as NCA, and providing tests and improved documentation to ensure code correctness and usability. Moreover, the user demonstrated expertise in adapting the code to newer versions of dependencies, and contributed towards single-sourcing the version number.
pythonmetric-learningmachine-learninglearning-algorithmsscikit-learn
scikit-learn/scikit-learn
Sep 2017 - Jul 2019
scikit-learn: machine learning in Python
Role in this project:Technical Writer
Technical WriterContributions:14 reviews, 20 commits, 29 PRs in 1 year 10 months
Contributions summary:William primarily contributed to the documentation of the scikit-learn library. Their commits focused on improving docstrings, fixing errors in existing documentation, and correcting links within the documentation to ensure accuracy and clarity. These changes involve updates to the documentation for specific classes and functions within the library, as well as overall improvements to the documentation structure.
data-analysispythonstatisticsdata-sciencelearn-machine-learning
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.