4 sources
A high-signal reinforcement-learning researcher whose work sits on the path from AlphaGo-era planning systems to Gemini-era reasoning and post-training techniques.
Researcher Profile
Editor reviewedRadu Soricut
Gemini (multimodal foundation models)
Distinguished scientist and senior research director at Google DeepMind
Important for understanding how multilingual NLP, translation, and multimodal reasoning meet inside production-scale frontier systems rather than staying separate research tracks.
Organizations
Google DeepMind
About This Page
This profile is meant to help you get oriented quickly: why this researcher matters, what to read first, and where to explore next.
Last reviewed
March 18, 2026
Known For
The ideas, systems, and research directions that make this person worth knowing.
01
Multilingual language systems
02
Vision-language models such as PaLI
03
Scaling language and multimodal research at Google
04
Gemini (multimodal foundation models)
05
Gemini: A Family of Highly Capable Multimodal Models
06
Gemini
Start Here
Canonical papers, project pages, or repositories that anchor this profile.
Signature Works
Additional papers, projects, or repositories that help flesh out the profile.
Supporting Sources
Additional links that help verify and flesh out this profile.
Related Researchers
People worth exploring next because they share topics, labs, or source material with this profile.
4 sources
One of the more useful people to study for the Gemini era because his work spans both the text-core of multimodal frontier models and the optimization tricks that make those systems cheaper and more stable to train.
4 sources
A high-signal researcher for understanding the modern scaling playbook, especially around compute-optimal training, retrieval-augmented language models, and the text side of Gemini-era multimodal systems.
4 sources
A good researcher to follow for the infrastructure side of frontier language models, especially mixture-of-experts scaling, instruction tuning, and the data systems that make very large models usable.
4 sources
Worth tracking for the data side of multimodal frontier models, where the quality and shape of training mixtures strongly determine what large systems can actually do.
4 sources
Important for the branch of DeepMind research that connects control, world models, and modern agent behavior rather than treating them as separate eras.