This talk presents a blueprint for bringing Dask workloads to HPC grids. We implement an architecture which allows dynamic sharing of compute resources within and between multi-tenant environments where Dask clusters are defined in secure, flexible, and repeatable templates.