Michel-Marie Maudet
Michel-Marie MAUDET is a leading figure in the Open Source community, co-founding and leading LINAGORA for over twenty years. This company stands as a beacon in the Open Source software and service industry, and is a major private investor in Open Source R&D across France, Vietnam, and Tunisia. LINAGORA offers a significant alternative to global tech giants by providing innovative solutions that empower businesses and governments towards digital sovereignty.
His career started in 1995 at the French Ministry of Defense, diving into systems engineering and discovering a deep passion for Open Source technologies in the early '90s. Michel-Marie has been dedicated to promoting the use of LINUX and Open Source, emphasizing the importance of open standards.
Recognizing the challenges and opportunities presented by Artificial Intelligence, LINAGORA launched LinTO, an Open Source personal assistant, in 2016. 2023, Michel-Marie further advanced his mission by founding the OpenLLM France (https://www.openllm-france.fr/) community, aiming to create a sovereign, transparent, and Open Source French LLM to challenge the dominance of major tech corporations and support the digital independence of France and Europe.
Michel-Marie’s vision extends beyond technology; he sees Open Source AI as a crucial tool for economic growth, job creation, and the competitiveness of strategic industries in Europe. His work champions a new path for digital innovation—one that is open, ethical, and responsible.
Session
At the FOSDEM '24 event, OpenLLM France evolved into OpenLLM Europe, embracing the goal of fostering an open collaborative space to develop partnerships among national initiatives aimed at creating open-source, sovereign, and efficient Large Language Models (LLMs) in the respective languages of each country.
OpenLLM Europe now boasts over 450 active members, receiving significant backing from the academic research community (22% of our members) and nearly 40% of our members are either users or companies specializing in AI.
Following the publication of CLAIRE in October 2023, our new model LUCIE has entered the pre-training phase. With 7B parameters, our primary objective is to address the underrepresentation of the French language in LLMs generally.
We have also incorporated other European languages, including German, Spanish, and likely Italian soon, alongside developing code to enhance the model's reasoning capabilities.
LUCIE is slated for 200,000 hours of pre-training on approximately 96 GPUs within the Jean ZAY machine, a sovereign supercomputer provided by our partner GENCI.
By June, LUCIE will be made available, having undergone transparent and independent evaluation by research teams from CNRS and INRIA.
This presentation will offer an update on the training progress, fine-tuning efforts, and the adoption of the LUCIE model.
Therefore, today in Europe, we have all the capabilities and, I believe with OpenLLM Europe, have generated the momentum, to build a real Open Source and sovereign European LLMs that meet our regulations.