Machine Learning Engineer @ LightOn
- 🚀 LightOnOCR (v1, v2, v3): state-of-the-art, efficient OCR models developed with large-scale training and RLVR. V3 comes in 0.8B, 1B, and 4B sizes.
- 🏗️ ModernBERT, contributed to architecture design, training and eval (ACL 2025)
- 🌐 ArabicWeb24, a 39B token Arabic corpus for LLM training
- 🛠️ vit.cpp, a lightweight C++ inference engine for Vision Transformers using GGML
- 💬 Interested in Optimization, Efficient Inference, GPU Kernels, Distributed Training, Reinforcement Learning (RL)
- 🎓 Engineering degree in maths and machine learning from École Centrale de Lyon
- 📫 Reach me: [email protected]
