executable jupyter notebooks
using docker and tmpnb
available: of

explore publicly available data sets

try new libraries like Thunder (built on Spark)

prototype analyses for benchmarking challenges