Baptiste Rozière
Baptiste Rozière is a computer scientist whose work connects machine translation with programming languages and AI coding assistants. He coauthored work on code translation using automated tests in 2021 and compiler representations in 2022. At Meta, he contributed to Llama and led Code Llama; in 2025, he led Mistral AI’s code generation team and coauthored the reasoning model Magistral. As of October 2026, he leads Mistral AI’s code generation team.
- Now at
- Mistral AI
- Works
- 42
Publication numbers from OpenAlex, as of 18 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.
Timeline
-
October 2021
Automated tests for code translation
In October 2021, he coauthors a method that uses automated unit tests to filter invalid code translations and build a tested parallel corpus.
-
June 2022
Compiler representations for translation
In June 2022, he coauthors work that uses compiler intermediate representations to improve code translation.
-
August 2023
Leading Code Llama
He leads Code Llama, a family of code models whose variants support infilling and instruction following; he also contributes to Llama.
-
June 2025
Magistral and reasoning models
In June 2025, he coauthors Magistral, which presents a reasoning model trained with reinforcement learning and describes Magistral Small as open source.
-
November 2025
Leading Mistral’s code team
In November 2025, he leads Mistral AI’s code generation team and gives a keynote on code assistants based on the team’s recent work.
-
As of October 2026
Scaling code models at Mistral
As of October 2026, he leads Mistral AI’s code generation team and is scheduled to discuss scaling code models.
Career
- Since 2025Mistral AICode generation team lead
- MetaCode Llama lead
Sources and credits
Photo: We have no published photo of Baptiste Rozière yet. Send us a free-licensed photo.
- Leveraging Automated Unit Tests for Unsupervised Code Translation
- Code Translation with Compiler Representations
- Code Llama: Open Foundation Models for Code
- From copilots to coworkers at AAAI: the gap between agentic research and production
- Magistral
- GHOST Day: Applied Machine Learning Conference
- Code assistants: from code completion to coding agents (AIware 2025 - Keynotes)
- Building Code Models: Scale and Beyond (SPLASH/ISSTA 2026 - REBASE industry track)
