← Back to Research

Automated Performance Tuning of ML Kernels for Heterogeneous Hardware

2025–Present • Collaborative Research

Working on an automated performance tuning framework for ML kernels in collaboration with Pacific Northwest National Laboratory (PNNL) and Stony Brook University, using LLM-based multi-agent approach with compiler tools like MLIR, IREE, and Triton. More updates coming up...

Keywords: Performance Tuning, ML Kernels, MLIR, IREE, Triton, Large Language Models, HPC

Collaborators: Pacific Northwest National Laboratory (PNNL) & Stony Brook University