How to provision different Python and R runtimes per user?

Currently, I have a Zeppelin box running in a mesos architecture talking to a Cloudera distro of HDFS using YARN, and we have several data scientists that use the Zeppelin box.

However, each data scientist has different needs/preferences regarding Python and R libraries. If I provision the Zeppelin box with a user that has permission to install any arbitrary Python and R library/package at some point the runtime will collide.

I would like to know if there is any best practice that is also scalable (meaning not having to provision a box per data scientist) to handle this situation appropriately.