It was good that you highlighted the wider ecosystem including Dask but also single node out-of-memory libraries like Vaex that offer some compatibility with the Pandas API.