A good BLAS implementation can improve the performance of many other libraries and platforms. A BLAS library is also the main back-end in deep learning frameworks which have been solving a ton of interesting problems in recent years. The main focus of this position is to measure and improve the performance of the SYCL-BLAS implementation, by contributing to the benchmark codes and adding missing features, such as other variants of GEMM operations or new hardware features such as sub-groups or other vendor-specific hardware features.