Raphael Sourty
Senior ML Engineer at LightOn
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolRaphael Sourty is a Senior ML Engineer based in Paris with nine years of experience building production-ready machine learning and NLP systems, now applying his expertise at LightOn. He holds a PhD focused on deep learning, language models and knowledge bases, and has translated that research into industry impact at ManoMano and Renault. A seasoned Python developer and Kaggle Expert, he contributes to open-source online-ML tooling—adding streaming statistics and imputers to the widely used river/creme modules—which highlights his attention to robust, real-time ML pipelines. Comfortable across research and applied settings, he blends rigorous statistical grounding with practical engineering to ship scalable information retrieval and language-model-driven solutions.
9 years of coding experience
3 years of employment as a software developer
DUT STID, Statistics, DUT STID, Statistics at Université de PoitiersPhD, Deep Learning, NLP, Knowledge Bases, Language Models, PhD, Deep Learning, NLP, Knowledge Bases, Language Models at Université Paul Sabatier Toulouse III
Github Skills (14)
incremental-learning10
statistics10
machine-learning10
online-machine-learning10
python10
streaming-data9
drift9
numpy8
data-structure7
data-science7
data-structures7
algorithm7
algorithms7
pandas6
Programming languages (6)
Github contributions (5)
online-ml/river
Feb 2019 - Jan 2022
🌊 Online machine learning in Python
Role in this project:
Data Scientist & ML Engineer
Data Scientist & ML EngineerContributions:2 reviews, 132 commits, 32 PRs in 2 years 11 months
Contributions summary:Raphael contributed significantly to the `creme/stats` module by adding new classes like `CentralMoments`, `Kurtosis`, and `Skew` to compute running statistics, specifically focusing on online learning algorithms. Their work involved implementing these new statistical features and modifying existing ones, and expanding the documentation, which suggests a focus on extending the library's capabilities for online machine learning. Furthermore, the addition of a numerical imputer indicates a willingness to improve the utility of the library with data preprocessing steps in a streaming context.
stream-processingpythonstreaming-datadata-scienceonline-statistics
raphaelsty/neural-cherche
Aug 2023 - Mar 2025
Neural Search
Contributions:17 releases, 5 reviews, 32 PRs in 1 year 7 months
colbertgooglelanguage-modelsemantic-searchsparseembed
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.