Specializing in high-performance computing pipelines, connecting the raw output of the ALMA array to our local tensor-processing nodes.