Publications

Sparse GPU kernels, workload characterization, and applied NLP, with Prof. David Kaeli at NUCAR.

IISWC2026

VCSR-TC: A Sparse Matrix-Matrix Optimization Strategy for Leveraging Modern GPUs with Tensor Cores

Mouad Tiahi, Elmira Karimi, David Kaeli

IEEE International Symposium on Workload CharacterizationBoulder, CO, Sep 2026

The new state-of-the-art SpMM format, feeding staged tiles straight to Tensor Cores for up to 4.44x over cuSPARSE.

ISPASS2026

VCSR-MM: A Bandwidth-Optimized SpMM Design for Modern GPUs

Mouad Tiahi, Elmira Karimi, David Kaeli

IEEE International Symposium on Performance Analysis of Systems and SoftwareSeoul, South Korea, Apr 2026

Bandwidth-optimized SpMM with stream tiling, beating cuSPARSE on A100, H100, and H200.

MIT URTC2024

Natural Language Processing in Environmental Health Research

Mouad Tiahi, Muskan Kumar, Zlatan Feric, David Kaeli

IEEE MIT Undergraduate Research Technology ConferenceCambridge, MA, Oct 2024

Optimized RAG for environmental health research.