Character-level MLP that generates Arabic names from scratch — built with PyTorch, inspired by Karpathy's makemore.
-
Updated
Aug 12, 2026 - Jupyter Notebook
Character-level MLP that generates Arabic names from scratch — built with PyTorch, inspired by Karpathy's makemore.
Offline Arabic name transliteration & fuzzy similarity. No LLM, no API calls, names never leave your infrastructure. Ships a 38 MB pruned XGBoost model + 22K EN↔AR dictionary. 98.4 lenient accuracy on MENA benchmark. Built for KYC, compliance, entity resolution, Arabic NLP preprocessing.
To associate your repository with the arabic-names topic, visit your repo's landing page and select "manage topics."