Introduction

Untitled

Hi! I am a newly minted Physics Ph.D. from Johns Hopkins University, interested in scientific computation, data science, and user experience research. In my Ph.D., I spent time thinking about processes governing the interactions of the tiniest constituents of nature, and the scientific tools physicists use to perform large scale particle physics experiments. My research was in collaboration with the Compact Muon Solenoid Experiment (CMS), a magnificent apparatus located at the Large Hadron Collider in CERN, Geneva. I studied exotic fundamental processes that solve unanswered puzzles in particle physics, using statistical modelling and scientific computation, with a special emphasis on using techniques from deep learning.

My time studying physics brought me a deep appreciation for principled scientific rigour and meticulous analysis. I am looking to pivot to fields where psychology, sociology and computer science meet, and I am extremely interested in human-computer interaction, empathy-driven technology and inclusive design.

In my free time I enjoy taking deep dives into dark comedy dramas, listening to progressive metal, lifting weights, and hosting mock talk shows for my close friends. I have a keen interest in good design, photography, and fine art, and dabble amateurly in all three.

Definitely reach out to me if anything on this page makes you curious for a chat.

[email protected]

github.com/SanjanaSekhar

linkedin.com/in/sanjanasekhar/

Resume

Academic CV

art_experiments

photoganj

Education

Ph.D. in Physics

Advisors: Morris Swartz, Petar Maksimovic

M.A. in Physics

M.Sc. in Physics + B.E.(Hons.) in Computer Science

Johns Hopkins University, Baltimore, USA

August 2021 - July 2025

Johns Hopkins University, Baltimore, USA

August 2019 - 2021

Birla Institute of Technology and Science, Goa, India

August 2013 - 2018

Research Projects

Search for the elusive Leptoquark particle using proton-proton collision data

The Compact Muon Solenoid experiment collects data from proton-proton collisions, which can then be analyzed using statistical tools in ROOT, C++ and Python. The major portion of my Ph.D. thesis has been spent performing a detailed statistical analysis to search for hypothetical particles named “leptoquarks”. Leptoquarks appear in several theoretical models that could explain open questions in the world of particle physics. Our result is available publicly on the CERN Document Server (CDS), and will be submitted to the Journal of High Energy Physics in October 2024. Read more about this project here:

Searching for virtual leptoquarks with the CMS detector

Paper on arXiv: arXiv:2503.20023

Github repo

Convolutional Neural Networks for position determination of charged particles

Particles produced in proton-proton collisions have to be detected with high precision in order to identify which physical processes were involved in the interactions. The CMS detector uses sophisticated electronics to reconstruct particle properties like position, mass, energy and momentum with excellent resolution. In this project, we built a fast convolutional neural network based algorithm to identify particle positions in the Pixel detector. Our algorithm improves upon existing position determination algorithms by 5% and is orders of magnitude faster. Read more about this project here:

Estimating particle positions in the CMS Pixel detector with CNNs

Github repo

Detecting anomalies in di-tau particle production using density estimation

To explain unanswered physics puzzles, theorists have produced thousands of models to be tested at particle colliders. Traditionally at CMS, an experimentalist chooses a model to perform a dedicated analysis. However this strategy is time-consuming and limited in person-power. New searches use model-agnostic anomaly detection techniques to exclude several theoretical models in the same analysis. We expand upon this idea in our project by looking for anomalies in proton-proton collisions that produce two tau particles in the final state. Using normalizing flows to estimate the distribution of background events, we classify events as signal like or background like in data. Read more about this project here: