Thomas Dybdahl Ahle

Hello, I'm Thomas. I am a Full Stack Research Scientist at Meta's "Mach­ine Lear­ning Efficiency" group, designing and implementing algorithms for scalable AI and Machine Learning. Pre­vi­ously I was the Chief of Machine Learning at the Natural Language Processing startup, SupWiz, and a Postdoctoral researcher in Theoretical Computer Sci­ence at the Basic Algorithms Research group (BARC) in Copenhagen with Mikkel Thorup. I did my PhD thesis with Rasmus Pagh on the Scalable Similarity Search project. Previous I worked with Oege de Moor and Samson Abramsky at the University of Oxford on Computational Ling­uistics; and with Eric Price at the University of Texas, Austin, on the fundamental limits of data-limited computation.

My research primarily involves scaling large transformer and recommendation systems, the theoretical foundations of machine learning and massive data, including similarity search, hashing, high dimensional geometry, kernel methods, sketching and derandomization.

Show more
Show more

Games

Teaching

Programming Problems

These are various algorithmic challenges I set on the Sphere online judge.

POWTOW, TRANSFER, REALROOT, NANO, WAYHOME, VFRIENDS, VFRIEND2

Media

Other

Contact

Coauthors