My work

Project: Journal Hub


Tech Stack

Summary

Journal Hub is a prototype online literature discussion platform, focused on publicly and privately curated collections of research publications.

Home page of the Journal Hub alpha.
Home page of the Journal Hub alpha
An article showcase in Journal Hub alpha.
An article showcase in Journal Hub alpha

As a platform, the vision of Journal Hub is to be an open space for discussing scientific research, and be a tool for research groups to collaborate on their own scientific scopes.

⚙️ Alpha development

Journal Hub is in its alpha development stage, however its core features have already been implemented. Along with a few close-circle volunteers, we are currently rolling out environments for interested groups, and we welcome all your feedback and ideas!

If you're interested in contributing or using Journal Hub, please send me an email: journal.hub.team@gmail.com

Research


Applying multivariate methods in Experimental Particle Physics Analysis

Libraries

Summary

During my doctorate research, I worked as an author of the ATLAS experiment under an analysis group focusing on a special category of particle interactions associated with di-boson (specifically $W^{\pm}Z$) production and scattering. The culmination of my thesis was to work with LHC data from the period between 2015 and 2018 to validate the observation of a purely electroweak signal and set experimental limits for New Physics, based on an extension of terms of the Standard Model Lagrangian of Particle Physics.

Likelihood-ratio for the observation of the purely electroweak $W^{\pm}Z$ signal
Likelihood-ratio for the observation of the purely electroweak $W^{\pm}Z$ signal

The statistical analysis compared single and multi-variate methods (traditionally Boosted Decision Trees) to an application of Neural Network trained utilizing a selection of topological features discriminating the various exepected theoretical signal signatures from the Standard Model background, with the resulting limits on novel interaction being far more constrained in the latter case. Limits on the New Physics (EFT) model parameters were extracted using the profile-likelihood ratio test statistic at a 68% and 95% C.L., being close to $0$ as no significant deviations in terms of previously unobserved interactions have been found.

Two-dimensional contour plot with limits on EFT model parameters
Two-dimensional contour plot with limits on EFT model parameters
High-level topological kinematic features used in Neural Network training
High-level topological kinematic features used in Neural Network training


You may find a full presentation of my thesis here!