Sudip Roy
Sudip Roy is a computer scientist whose work centers on the systems that train and serve large machine-learning models. He holds a B.Tech from IIT Kharagpur and finished a PhD at Cornell in 2014; a paper he co-authored there on entangled queries won the SIGMOD 2011 best paper award. At Google Brain he worked on systems research and contributed to TensorFlow Extended and Pathways, and the Pathways paper won an MLSys Outstanding Paper Award in 2022. After rising from engineering manager to senior director of inference at Cohere between 2021 and 2025, he co-founded Adaption Labs with Sara Hooker in 2025, and the company raised a $50 million seed round in 2026. He is co-founder and chief technology officer of Adaption.
- Now at
- Adaption Labs
- Works
- 26
- Citations
- 1,473
- h-index
- 14
Publication numbers from OpenAlex, as of 22 October 2026.
Timeline
-
2011
A SIGMOD best paper
He is a co-author, with Cornell colleagues including Johannes Gehrke, of "Entangled queries: enabling declarative data-driven coordination", which wins the SIGMOD 2011 Best Paper Award.
-
By 2014
A Cornell PhD
He holds a B.Tech in computer science and engineering from the Indian Institute of Technology Kharagpur. He finishes a PhD in computer science at Cornell under Johannes Gehrke in 2014.
-
By 2021
Systems work at Google Brain
At Google Brain, he works on systems research and large-scale data management, and contributes to infrastructure projects including TensorFlow Extended and Pathways. In 2021, he is a co-author of an MLSys paper that learns a performance model for Tensor Processing Units from a corpus of tensor computation graph programs.
-
2022
Pathways at MLSys 2022
He is a co-author of Pathways: Asynchronous Distributed Dataflow for ML, which presents an orchestration layer for accelerators designed to enable exploration of new systems and ML research ideas. The paper receives an Outstanding Paper Award at MLSys 2022.
-
2021–2025
Engineering leadership at Cohere
He joins Cohere as an engineering manager in November 2021 and becomes director of engineering in June 2023. From January 2025, he is senior director of inference, leading work on inference infrastructure and fine-tuning systems.
-
October 2025
Co-founding Adaption Labs
In October 2025, he joins Sara Hooker, the former head of Cohere Labs, in launching Adaption Labs. The startup aims to build efficient AI that can adapt and continuously learn.
-
February 2026
A $50 million seed round
Adaption Labs raises $50 million in seed funding in February 2026. Emergence Capital Partners leads the round, with Mozilla Ventures among the participants.
-
As of October 2026
Chief technology officer of Adaption
He is co-founder and chief technology officer of Adaption, a San Francisco company building AI that adapts and keeps learning. In May 2026, Adaption launches AutoScientist, an automated fine-tuning system that the company says raised win rates from 48% to 64% against configurations its own AI researchers picked.
Career
- Since 2025Adaption LabsCo-founder, Co-founder and Chief Technology Officer
- 2021–2025CohereEngineering Manager, Director of Engineering
- 2015–2021Google
Education
- 2014Cornell UniversityPhD in computer science
- Indian Institute of Technology KharagpurB.Tech in computer science and engineering
Sources and credits
Photo: We have no published photo of Sudip Roy yet. Send us a free-licensed photo.
- SIGMOD 2011 Conference: top presence for IC researchers - EPFL
- Sudip Roy | Torre.ai
- Sudip Roy (computer scientist) - Wikipedia
- [2008.01040] A Learned Performance Model for Tensor Processing Units
- Track: Outstanding Papers - MLSys 2022
- [2203.12533] Pathways: Asynchronous Distributed Dataflow for ML
- Ex-Cohere execs Sara Hooker and Sudip Roy unveil new AI startup | BetaKit
- Why Cohere's ex-AI research lead is betting against the scaling race | TechCrunch
- Adaption Labs secures $50 million seed round to build AI models that can change on the fly | Fortune
- Sara Hooker's Adaption launches AutoScientist with $50M



