Noam Shazeer
AI Scientist at OpenAI
A researcher behind attention architectures, mixture-of-experts models, and the systems used to train large language networks.
Biography
Noam Shazeer is an American computer scientist whose career spans language-model research, large-scale computing, and AI entrepreneurship. He studied at Duke University and joined Google in 2000. His early work included improvements to search spelling correction and advertising systems, before his research became closely associated with neural language technology.
Shazeer co-authored Attention Is All You Need in 2017 and contributed to sparsely gated mixture-of-experts research. These projects explored architectures and ways to increase model capacity while controlling computational cost. His work also included Mesh-TensorFlow, which addressed how very large models could be distributed across computing hardware.
In 2021, he co-founded Character.AI with Daniel De Freitas, developing conversational systems around interactive characters. He later returned to Google and co-led work on Gemini. That career path connected fundamental model research with both consumer-facing AI products and industrial programs operating at substantial scale.
Shazeer's personal website identifies him as working at OpenAI following his move in 2026. His contribution to AI includes the design of influential components as well as the engineering needed to make large models trainable. His history also illustrates the close relationship between research architectures, computing systems, and the forms of conversation that reach end users.
