1 articles with this tag
Morph's Tejas Bhakta told AI Engineer how an autoresearch loop tuning CUDA kernels delivers 3x inference speedups, with humans supplying ideas and agents searching parameters.
