creators

Tri Dao: FlashAttention, Mamba, Together AI

Tri Dao, co-founder and chief scientist of Together AI and a Princeton computer scientist, created FlashAttention and co-created the Mamba architecture.

Tri Dao works at the intersection of machine learning and systems. His CV lists him as co-founder and chief scientist of Together AI since July 2023. Princeton's computer science department lists him as a visiting research scholar and says he was previously an assistant professor; his own CV, site and X bio still carry the assistant-professor title, dated from September 2024, and his site says he leads the Dao AI Lab. He took a BS in mathematics and two master's degrees at Stanford, and a PhD in computer science there in 2023, advised by Christopher Re and Stefano Ermon. He is known for FlashAttention, the IO-aware exact attention algorithm published in 2022, and its successors up to FlashAttention-4, a best paper honourable mention at MLSys 2026. With Albert Gu he wrote Mamba in 2023, a selective state-space architecture offered as an alternative to transformers; it won an outstanding paper award at COLM 2024. His Monarch paper was the ICML 2022 outstanding paper runner-up, and Schmidt Sciences named him an AI2050 early career fellow in November 2025.

Specialization

efficient attention algorithms, state space models, ML systems

Country

United States