Julien Plu
Research Scientist at Lettria
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolJulien Plu is a research scientist in Paris with 11 years of experience building production-ready NLP and ML systems, currently focusing on graph engineering for recommendation, GraphRAG and integrating graphs into LLM reasoning. He holds a PhD in Informatics and has a strong track record of moving research into production—from training French language models at Leboncoin to optimizing compact, low-latency LLMs at Buster.AI and on-device inference work at Beam. Julien has been an active contributor to Hugging Face’s flagship projects, implementing the TensorFlow trainer for transformers and enhancing the datasets library, showing deep familiarity with model tooling, multi-GPU/TPU training and data pipelines. Comfortable across the full stack (Python, C++ integrations, GCP), he specializes in adaptive entity linking, semantic web approaches and practical ML engineering at scale. Outside work he channels the same discipline into decade-spanning pursuits—competitive tennis, 20+ years of scuba diving and playing hard rock guitar—reflecting a blend of technical rigor and adventurous curiosity.
11 years of coding experience
12 years of employment as a software developer
Bachelor of Applied Science (B.A.Sc.) computer science, Bachelor of Applied Science (B.A.Sc.) computer science at University of Montpellierhigh school graduation Engineer Science option and Physics speciality, high school graduation Engineer Science option and Physics speciality at High School Joliot Curie (Sète, France)Doctor of Philosophy - PhD Informatique, Doctor of Philosophy - PhD Informatique at Télécom ParisMaster's degree Informatics, Master's degree Informatics at Université Montpellier 2
French, English, Spanish, Italian
Github Skills (12)
pandas10
machine-learning10
nlp10
deep-learning10
tensorflow10
python10
datasets10
data-pipelines9
modeling9
data-pipeline9
pytorch8
numpy8
Programming languages (14)
Github contributions (5)
huggingface/transformers
May 2020 - Jan 2022
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:
Back-end & ML Engineer
Back-end & ML EngineerContributions:366 reviews, 114 commits, 195 PRs in 1 year 9 months
Contributions summary:Julien primarily contributed to developing a TensorFlow (TF) version of a trainer for the Hugging Face Transformers library. Their work included the initial implementation of the TF trainer, refactoring code for compatibility with the PyTorch trainer, and adding examples such as a question-answering (QA) trainer. The user also addressed bugs related to prediction, multi-GPU training, and optimization tests, while also adding support for new features such as TPU strategy.
pythonbertspeech-recognitionstate-of-the-artflax
huggingface/datasets
Apr 2020 - Dec 2020
🤗 The largest hub of ready-to-use datasets for ML models with fast, easy-to-use and efficient data manipulation tools
Role in this project:Back-end Developer & MLOps Engineer
Back-end Developer & MLOps EngineerContributions:8 reviews, 23 commits, 26 PRs in 7 months
Contributions summary:Julien contributed to bug fixes and feature enhancements related to the `datasets` library. Their work involved modifying code in `arrow_reader.py`, `info.py`, and `arrow_dataset.py`, indicating familiarity with the internal workings of the library and data processing pipelines. The user also added support for "pandas" as a format type within the dataset, and integrated CSV datasets, which points to a strong involvement in enhancing the data handling and format capabilities of the library. Finally, the user appears to have updated to TF.ragged.constant, which is related to model integrations.
ml-modelstensorflownatural-language-processingmanipulationdata-science
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.