I am most interested in in the new pipelines for MLLib and the SQL interface.
There are a few notebooks with my worked examples of querying data loaded from a CSV file using the dataframe-API. I also train a few models using the pipeline framework. It's pretty slick.
You can run the code in a docker image I got from here. I volumed in the contents of the work folder to the /home/jovyan directory. that allows your notebooks and data. You can use the command:
docker run ezamir/jupyter-spark-2.0 -p 8888:8888 -v path/to/your/host/directory:/home/jovyan/work