The high energy physics community at the LHC have utilized state of the art machine learning (ML) techniques quite successfully including the discovery of the Higgs boson. While such methods are quite common and readily applicable in big data style analysis, they are only recently being formulated and employed in nuclear physics. We shall go through a very brief overview of machine learning techniques and some of its recent applications to the study of the Quark Gluon Plasma. We study the phenomenon of jet quenching utilizing quark and gluon jet substructures as independent probes of heavy ion collisions. We exploit jet and sub-jet features to highlight differences between quark and gluon jets in vacuum and in a medium with the jet-quenching model implemented in JEWEL MC. To systematically extract jet substructure information, we introduce the telescoping deconstruction framework exploiting subjet kinematics at multiple angular scales. We find that the quark gluon discrimination performance worsens in heavy ion jets due to significant soft event activity affecting the soft jet substructure. Our work suggests a systematically improvable framework for studying modifications to quark and gluon jet substructures and facilitating direct comparisons between theoretical calculations, simulations and measurements in heavy ion collisions.