Recent Publication at IPDPS ’26

A big shot out to Dr. Devarakonda!

His paper, co-authored with researchers at Cornell University and the University of Trento entitled “Communication-Avoiding Linear Algebraic Kernel K-Means on GPUs” has been published at the 40th IEEE International Parallel and Distributed Processing Symposium (IPDPS ’26)!

The paper tackles the problem of grouping large collections of data into meaningful clusters. The team designed new algorithms that spread the work across many GPUs while keeping communication between them to a minimum. The result scales to datasets 10 to 100 times larger than was previously practical and cuts clustering time from over an hour to under two seconds using up to 256 GPUs.

To read more, please visit HERE.