Etienne Bernard builds specialized language models for information extraction at NuMind. He focuses on solving the hallucination and layout issues that cause generic models to struggle with complex documents.
Before founding NuMind, he led the machine learning team at Wolfram Research and wrote Introduction to Machine Learning. Today, his work centers on open-source tools like NuExtract and NuMarkdown, proving that smaller, targeted models can outperform giants like GPT-5 in specific extraction tasks.
He holds a PhD in physics from ENS and MIT.
