Philomène Chagniot

Timeline

1. 2020

   Engineering degree at ENSAE Paris

   She is an ENSAE Paris engineer of the class of 2020, with an MSc in Data Science and a master's in Mathematics, Vision and Learning (MVA) from ENS Paris-Saclay. Earlier degrees include a Bachelor of Arts in economics from Sciences Po and a Bachelor of Mathematics from Université Pierre et Marie Curie. Internships on her path include natural language processing at the OECD and recommendation-systems research at Criteo and ENS Paris-Saclay.

2. September 2020

   Long-term rewards in RecoGym

   In September 2020, she co-authors "From Clicks to Conversions: Recommendation for long-term reward" with Flavian Vasile and David Rohde. The paper introduces a framework for modeling long-term rewards in the RecoGym simulation environment and proposes a simple extension for conversion-optimized recommendations.

3. March 2024

   CLIP for few-shot classification

   From InstaDeep in Paris, where she works as an AI research engineer, she co-authors "Multimodal CLIP Inference for Meta-Few-Shot Image Classification" with Constance Ferragu and Vincent Coyette in March 2024. The paper shows that combining modalities from CLIP's text and image encoders outperforms state-of-the-art meta-few-shot learners on widely adopted benchmarks without additional training.

4. June 2025

   Co-authoring Magistral

   In June 2025, she is among the authors of Magistral, Mistral's first reasoning model and its own scalable reinforcement learning pipeline. The paper presents Magistral Medium, trained for reasoning with reinforcement learning alone, and the open-source Magistral Small.

5. July 2025

   Co-authoring Voxtral

   She is a co-author of Voxtral Mini and Voxtral Small, two multimodal audio chat models that comprehend spoken audio and text documents, in July 2025. Both models are released under the Apache 2.0 license, and Voxtral Small outperforms a number of closed-source models while being small enough to run locally.

6. January 2026

   The Ministral 3 series

   Her name appears on Ministral 3 in January 2026, a family of parameter-efficient dense language models with 3B, 8B and 14B parameters. Each size comes as a pretrained base model, an instruction-finetuned model and a reasoning model, all with image understanding under the Apache 2.0 license.

7. July 2026

   Shieldstral and Robostral Navigate

   In July 2026, she joins the author lists of two papers: Shieldstral, a 3B-parameter multimodal safety classifier, and Robostral Navigate, an 8B vision-language model that predicts navigation waypoints from monocular RGB images.

8. As of October 2026

   AI scientist at Mistral AI

   ENSAE Alumni announces in October 2025 her appointment as AI Scientist at Mistral AI. Her earlier roles include AI research engineer at InstaDeep.