I subdivide my main research topics into three broad areas in this page. On a practical matter, on a day-to-day basis, there are often several degrees of overlap between them but for simplicity, I will list them individually here as three questions.  

The topics presented below are aimed at the general level which is why I dont provide details. For a more technical summary of these topics and the latest measurements, please take a look at some of the publications or my talks and/or feel free to contact me :) 

How do quarks and gluons become hadrons? 

One of the cornerstone properties of fundamental particles such as quarks or gluons that carry color charge and interact via the strong force is that they are never found in isolation. They are always found as color neutral hadrons in nature and this has inturn given rise to one of the main unsolved questions (that has always captivated me) - how do color charged partons (quarks/gluons) end up as color neutral hadrons. The curious nature of high energy collisions, proton-proton collisions for example, is that while one can experimentally measure the final-state particles and also understand that the initial interactions happened between the partons, we dont know how exactly the evolution happens. This transition from perturbative (calculable interactions/emissions of quarks and gluons) to non-perturbative (models based on fits to measurements) is what I would like to experimentally measure. Such measurements not only provide an experimental handle on how the quarks and gluons evolve, but simultaneously also answer the question of when the transitions occur which can be used to develop accurate modeling of non-perturbative physics with the hope that ultimately, we can truly understand the processes leading to the formation of hadrons. 

What is the space-time structure of the Quark-Gluon Plasma?

The quark-gluon plasma (QGP) is an asymptotically free phase of matter composed of de-confined quarks and gluons, theorized to have existed in the early universe, a few microseconds after the big bang. Collisions of heavy ions such as gold (at RHIC, amongst other species) or lead (at LHC) at relativistic speeds produce 'little-bangs' which have the right initial conditions to produce the QGP for a fleeting moment. The particles produced in these collisions have been shown to behave in a manner reproducible with hydrodynamic calculations of a relativistic fluid with non-zero shear and bulk viscosity. Since the QGP phase formed after the collision, exists at an extremely high temperature and for a very short moment in time, one cannot study it with external probes. An experimental approach to study this phase of matter and its transport properties is to utilize internal probes, which are high energy quarks or gluons that are produced during the initial collision and traverse the hot and dense medium. I mentioned in the previous section, the techniques we use to study the evolution of the quarks and gluons in proton-proton collisions can be similarly employed in these heavy ion collisions to perform a comparative study wherein the modifications can point towards understanding of the medium's transport properties. Some of my recent work in this area aims utilize these techniques to expose the space-time evolution of the QGP with the use of differential measurements from the STAR collaborations. 

How can machine learning extend the boundaries of current measurements? 

Machine learning (ML) in the context of high energy collider physics can be operationally defined as an area of study where algorithms are used to represent, learn and extract information from multi-dimensional data. These datasets can be anything from low level detector information or higher level analysis objects such as reconstructed particles or clusters. This topic had already found a home in experimental collider physics in proton-proton collisions with it being instrumental in the discovery of the Higgs particle. Measurements in heavy ion physics suffer from both correlated and uncorrelated background where it is practically an ill-defined problem to identify one vs the other. Appropriately defined and trained ML models offer an exciting opportunity to exploit hitherto unknown correlations within higher dimension data. This can be both at the reconstruction level, such as identifying or categorizing special types of events or at the analysis level where image or graph style network offer an advantage over current methods.