Patrick Pérez

Also known as Patrick Perez

Patrick Pérez is a Paris-based computer scientist who has worked on computer vision and machine learning in public research, industry labs and a non-profit AI lab. After a doctorate at the Université de Rennes in 1993, he worked at Inria, Microsoft Research Cambridge and Technicolor, and co-authored the 2003 SIGGRAPH paper "Poisson Image Editing" with Andrew Blake and Michel Gangnet. From 2018 to 2023, he was VP of AI and Scientific Director of valeo.ai, Valeo's Paris research center for artificial intelligence. In 2023, he became CEO of the non-profit open-science lab Kyutai; in 2024, he co-authored Moshi, a speech-text foundation model. He is Research Director at Google DeepMind in Paris.

Now at: Google DeepMind
Born: 1968
Badges: Founder
Bridge

Years link to their sources.

Timeline

1993

Doctorate at Rennes

After graduating from École Centrale Paris in applied mathematics, he defends a doctoral thesis at the Université de Rennes in 1993 on Markov fields and multiresolution analysis of images, applied to motion analysis. He also works as a postdoc at Brown University.
2000–2004

Microsoft Research Cambridge

In 2000, he leaves Inria for Microsoft Research Cambridge, where he works as a research scientist until 2004.
2003

Poisson image editing

In 2003, he co-authors "Poisson Image Editing" for SIGGRAPH with Andrew Blake and Michel Gangnet. The paper solves Poisson equations to import opaque and transparent source regions into a destination image seamlessly and to modify the texture, illumination or color of a selected region.
2004–2018

Inria and Technicolor

He is back at Inria from 2004 to 2009, and his career there includes a post as research director. In 2009, he moves to Technicolor, where he works as a research scientist until 2018.
2018–2023

Scientific Director of valeo.ai

In January 2018, Valeo appoints him Scientific Director of valeo.ai, its Paris research center for artificial intelligence and deep learning in automotive applications. He holds the title of VP of AI and Scientific Director until 2023 and co-authors ADVENT in 2019, a CVPR paper on unsupervised domain adaptation for semantic segmentation.
November 2023

CEO of Kyutai

Kyutai launches at Station F in Paris in November 2023 as a non-profit open-science AI lab founded by the iliad Group, the CMA CGM Group and Schmidt Futures. He leads the lab as CEO, with more than thirty years of experience in computer vision and machine learning.
2024–2025

Moshi and Helium-1

In 2024, he is one of eight authors of Moshi, a speech-text foundation model and full-duplex spoken dialogue framework. In January 2025, the Kyutai team under his direction unveils Helium-1, a multilingual text model with two billion parameters built to run locally on a computer or phone.
As of October 2026

Research Director at Google DeepMind

As of October 2026, he is Research Director at Google DeepMind in Paris, after serving as co-founding CEO of Kyutai.