Broadly, this group performs electrophysiologic studies of decision-making brain networks, in rodents and in humans. In those experiments, the researchers routinely record 100-300 channels of continuous, 30+ kHz data for several hours. They then decompose those data into time-frequency and time-frequency-connectivity representations and test for differential activity between experimental conditions. The preferred statistical approach in that scenario is non-parametric permutation testing, with 1,000 recomputations of the same analysis on shuffled versions of the original data. These are computationally expensive MATLAB/Python analyses. They are also extremely parallelizable, and the researchers are using MSI resources to perform that parallel computation.