At a time when large language models occupy a central place in the ecosystem of artificial intelligence, some actors choose to develop complementary technologies, focused on reading, structuring and analysis of large -scale data. It is the positioning of Arlequin Ai, a young French company specializing in the analysis of digital social dynamics. Its technology is based on a non -generative approach, based on non -supervised artificial intelligence algorithms.
Founded in 2024 by Hugo Micheron, researcher in geopolitics and social sciences, and Antoine Jardin, research engineer at CNRS and specialist in data science, Arlequin AI develops tools intended to restore a clear and exploitable vision of complex and heterogeneous data. Their main solution, Hudex (Human Data Explorer), makes it possible to detect, organize and interpret social and narrative dynamics from raw corpus, whether textual, visual or sound.
A technology built to understand rather than produce
Arlequin Ai is not intended to generate content or simulate human interaction, as is the case with LLM -based interfaces (large Models). The startup wants to provide analysts, decision -makers and tools to structure, understand and follow massive information flows.
Hudex can process multilingual data, structured or not, and restore a synthetic reading of narrative structures and emerging signals. The operation is based on an analysis without requests where the user imports his data, and the tool performs full processing, without depending on a pre-trained model or a prompt. This approach aims to limit biases, while guaranteeing the reproducibility and auditability of the results.
Concrete applications and areas of use
The technology developed by Arlequin AI is already used in several professional contexts. It was notably deployed to ministries, private groups and public media like Radio France. In June 2025, it allowed the massive processing of online data from the great national debate, demonstrating its ability to structure large -scale public discussions.
Beyond the public sector, Arlequin AI also offers cases of use in finance (detection of weak signals, regulatory watch), logistics (anticipation of breaks), management of reputational risks, or the fight against disinformation.
A modular and sovereign technological model
On the technical level, Arlequin Ai has chosen a modular architecture, based on specialized open source components, which it assembles and adapts internally. Algorithms are designed to operate on both GPUs and CPUs, allowing on -site deployments in sensitive environments.
The company claims complete autonomy on its technological pipeline, without dependence on external models or to non -European cloud platforms. This choice aims to meet the growing requirements of digital sovereignty and guarantee the confidentiality of the data processed.
Funding to strengthen the industrialization of the solution
In order to support the industrialization of its tools, Arlequin ai announces a lifting of 4.4 million euros in priming. This tour of the table is led by Vsquared venturesGerman fund specializing in Deeptech. It also brings together 10x Founders,, Kima Ventures,, Beetleas well as several French private investors, including David Meneret,, Arthur Dénouveaux,, Éric Le Maire,, Julien Codorniou And Julien Lesaicherre.
This fundraising will allow the company to Double its R&D workforceto strengthen its on-premise deployment capacities, and to expand its presence on a European scale. Arlequin has currently had around twenty employees from both social sciences, IT, and engineering of complex systems.
