Wu Yonghui

Overview

Wu Yonghui is an artificial intelligence researcher who spent nearly two decades at Google and now works at ByteDance. At Google he worked on the Google Brain deep-learning team from 2014 and was promoted to vice president of research at Google DeepMind in September 2023^[1]. In February 2025 he reportedly joined ByteDance, where he heads foundational research on large models^[1][2][3].

At Google

Wu spent nearly two decades at Google and holds the title of Google Fellow^[1]. Fellow is a senior technical rank that large technology companies use to mark scientists and engineers at the top of the technical career ladder — a track that runs alongside management rather than through it.

From 2014 he worked on Google Brain, Google's deep-learning research team^[1]. Deep learning is the approach in which neural networks with many layers are trained on large amounts of data so that the network learns useful representations on its own, instead of having features hand-designed by people. A dedicated team of this kind works on the methods themselves and on the training systems that support them.

In September 2023 he was promoted to vice president of research at Google DeepMind^[1]. A vice president of research in an AI organisation typically sets the direction of a research portfolio and leads the scientists and engineers who carry it out, sitting between senior leadership and individual research teams.

Move to ByteDance

Wu reportedly joined ByteDance on 17 February 2025^[1]. He joined the company's Seed large-model team as head of foundational research, overseeing foundational infrastructure and algorithm design for large models in a role similar to that of a chief scientist^[2][3].

The two halves of that brief map onto the two things a large-model effort needs. Foundational infrastructure covers the computing clusters, data pipelines and training frameworks a model is built on; algorithm design covers the model architecture and the training methods themselves. Putting both under one person means the research agenda and the machinery that supports it are set together.

He reports directly to ByteDance CEO Liang Rubo^[2]. After his arrival, several algorithm and technology leaders who had reported to Seed head Zhu Wenjia began reporting to him^[2].

The Seed team

Zhu Wenjia joined ByteDance in 2015 and led the creation of the company's first large language model team in 2023, the same year Wu was promoted at Google DeepMind^[1]. Zhu had previously been global research and development chief for TikTok and chief executive of Toutiao^[1].

Why the move was noticed

Large-model work is expensive and depends on a small number of people who can both set a research direction and judge whether the engineering behind it will scale. Moves of senior researchers between major labs are therefore read as signals about where that kind of capacity is being built. In this case the reporting lines are part of the story: algorithm and technology leaders moving under the new head of foundational research is a concrete sign of how the team's structure was rearranged around foundational work^[2].