Repo (my supervisor’s GitHub): Github – BatchedKernels.jl My master’s project focuses on accelerating linear algebra operations on large batches of small matrices using GPUs. This implementation outperforms state-of-the-art libraries such as JAX by as much as 30x. Overview # Modern GPU-accelerated libraries are primarily optimised for large matrices, as large matrices typically come with larger workloads, and GPU parallelisation is more necessary.

Read it here.