·
Builders & Technical Leaders
Lysandre Debut
Chief Open Source Officer |Hugging Face
Core maintainer of Transformers, led the v5 release. Co-authored the EMNLP 2020 Transformers paper. First engineer at HF focused entirely on open source.
1papers
GitHubLysandreJikResearch Papers
1 paperBiography
Deep Research
Lysandre Debut is the Chief Open Source Officer at Hugging Face and a core maintainer of the Transformers library, the most widely used framework for state-of-the-art machine learning models in text, vision, audio, and multimodal domains. He co-authored the seminal paper "Transformers: State-of-the-Art Natural Language Processing" presented at EMNLP 2020, and was the first engineer at Hugging Face to focus entirely on the open-source mission. Educated at ESEO Angers in Computer Science Engineering with a Data Science major (2016-2019), he joined Hugging Face in 2019 and has since led the growth of the Transformers library from 40 architectures and 20,000 daily downloads to over 400 architectures and 3 million daily pip installs. In January 2026, he shipped Transformers v5 -- the first major version release in five years -- featuring performance optimizations, a unified tokenizer backend, PyTorch-first focus, and deep ecosystem interoperability.
Transformer ArchitecturesOpen Source ML InfrastructureNatural Language ProcessingModel InteroperabilityTokenizationSafetensorsDiffusion ModelsPyTorch EcosystemModel Hubs and DistributionML Benchmarking
Timeline
10 Research10 total
2026
2025
2023
2020
2019
2017
2017-08Research
Created GitHub account (LysandreJik)Key Contributions
Hugging Face Transformers
Core maintainer and co-author of the Transformers library, the de facto standard for state-of-the-art ML models. Over 400 architectures, 750,000+ model checkpoints on the Hub, and 3 million daily pip installs. Led the Transformers v5 release.
Transformers v5
Led the first major version release in five years. Introduced modular model definitions, unified tokenizer backend (removing slow/fast split), PyTorch-first backend, AttentionInterface abstraction, and deep interoperability with vLLM, SGLang, llama.cpp, MLX, and TensorRT.
EMNLP 2020 Paper: Transformers Library
Co-authored the foundational paper "Transformers: State-of-the-Art Natural Language Processing" (Wolf, Debut et al.), published at EMNLP 2020 and cited thousands of times.
Safetensors Adoption
Drove the default adoption of safetensors serialization in Transformers (v4.35+), improving model loading safety and performance by replacing pickle-based torch files.
Hugging Face Diffusers
Contributed to the Diffusers library for state-of-the-art pretrained diffusion models, enabling generation of images, audio, and 3D structures.
Benchmarking Transformers
Published benchmarking analyses comparing PyTorch and TensorFlow performance for Transformer inference, guiding framework optimization decisions.
10 sources(click to expand)
LysandreJik GitHub ProfileLysandre on Hugging Face HubTransformers: State-of-the-Art Natural Language Processing (EMNLP 2020)arXiv: HuggingFace's Transformers (1910.03771)Transformers v5: Simple model definitions powering the AI ecosystem (HF Blog)Lysandre Debut LinkedIn ProfileLysandre's Transformers v5 RC announcement on XBenchmarking Transformers: PyTorch and TensorFlow (Medium)PyTorch Conference 2023 — Diffusers Lightning TalkLysandre Debut on Google Scholar
Research generated March 19, 2026
Builders & Technical Leaders/Lysandre Debut
All Profiles