Rajagopal Venkat

Computer Science Ph.D. Candidate, Washington University in St. Louis
rajagopal@wustl.edu

I am a Ph.D. candidate at Washington University in St. Louis, advised by Dr. Yevgeniy Vorobeychik. My research interests are genetic privacy, adversarial machine learning, complex network models and natural language processing. My work in genetic privacy is in collaboration with the GetPreCiSe center.

"There are lies, there are damned lies and there are statistics." Mark Twain, quoting somebody else.
"Never trust anything that can think for itself if you can’t see where it keeps its brain." Arthur Weasley


Education

Washington University in St. Louis

Ph.D.

Computer Science

Advisor : Dr. Yevgeniy Vorobeychik

August 2018 - Present

Washington University in St. Louis

M.S.

Computer Science
Granted Jan 2021

Vanderbilt University

Ph.D. (partial coursework)

Computer Science

Advisor : Dr. Yevgeniy Vorobeychik

August 2017 - August 2018

Shiv Nadar University

Bachelor of Technology

Major: Computer Science

Minor: Mathematics (emph. Graph Theory, Complex Networks)

August 2013 - May 2017

Research

I have been fortunate enough to work on a variety of problems spanning several research domains during my PhD. My primary interests are in Data Privacy, Machine Learning and AI, Graph Theory, and NLP. Please do get in touch with me, should any of my work pique your interest.


Select Projects


Preserving Privacy in Genomic Summary Information Releases
We present rigorous mathematical frameworks to guarantee privacy against membership inference attacks when genomic data are released in the form of summary statistics such as alternate allele frequencies or binary presence/absence queries. We develop highly-scalable heuristics that enable the data custodian to finely-tune the utility-privacy balance of such data releases.

Matching face images to DNA
We evaluated the risk of reidentification in a database of SNPs based on individuals’ facial images in the wild. We used deep neural networks to learn phenotypes (facial features) from images, then probabilistically match them to genomes by predicting phenotypes from the latter. To mitigate such an attack, we constructed adversarial examples using gradient-based methods.

Topic Modeling
To overcome certain shortcomings of probabilistic topic modeling on short-text datasets, we developed a deterministic topic modeling method that leverages semantic similarity and word embeddings.

Community Detection
We developed a scalable community-detection algorithm for very large weighted, directed real-life networks based on multiple simulated parallel random walks.


Publications/Preprints


Defending Against Membership Inference Attacks on Beacon Services Rajagopal Venkatesaramani, Zhiyu Wan, Bradley A. Malin and Yevgeniy Vorobeychik. Arxiv [cs.CR] 2112.13301, 2021

Re-Identification of Individuals in Genomic Databases using Public Face Images Rajagopal Venkatesaramani, Bradley Malin and Yevgeniy Vorobeychik. Science Advances, Vol 7, Issue 47, 17 Nov 2021. DOI: 10.1126/sciadv.abg3296

A Semantic Cover Approach to Topic Modeling. Rajagopal Venkatesaramani, Douglas Downey, Bradley Malin and Yevgeniy Vorobeychik. Proceedings of the Eighth Joint Conference on Lexical and Computational Semantics (*SEM), pages 92-102

Community Detection by Information Flow Simulation Rajagopal Venkatesaramani and Yevgeniy Vorobeychik. Arxiv [cs.SI] 1805.04920, 2018

Personal

TLDR; Musician | Literature Aficionado | Coffee Snob | Petrolhead | Aviation Enthusiast


My story begins in the tiny town of Tenkasi, Tamil Nadu in India. Soon after I was born, we moved to the capital city of New Delhi for better work opportunities presented themselves to both my parents - my father chose to pursue accounts despite a background in Physics, and my mother found joy in teaching Mathematics. However, during my school years, I was anything but studious. Not that I did not enjoy my subjects - I had exceptional teachers - but I enjoyed not being in class (especially without official blessing) far more. Instead, I would spend my time indulging in every extra-curricular activity the school had to offer, or hiding in the music room practicing in a hard-to-spot corner, much to the annoyance of my teachers. Throughout school, I would repeatedly quote the words of Rear Admiral Grace Hopper, "If it's a good idea, go ahead and do it. It's much easier to apologize than to seek permission." Little did my family of educators expect me to end up becoming one.

Growing up in Delhi is an aspect of my life I'm deeply grateful for; the city is a melting pot of cultures - cultures spanning 1.3 billion people. The result is an endearing sort of mess - nothing works the way it is supposed to, nothing is where you'd expect it to be, nothing is predictable, and yet life is perfectly normal for the average dweller. Ordo ab Chao. Residents of Delhi thus tend to be quite creative with everyday solutions, and function on what is a rather untranslatable term, jugaad - use only what you have, but get the job done anyway. It is an attitude one develops over time, as Delhi holds your hand and teaches you fortitude.

The four years of my undergraduate degree at Shiv Nadar University were truly transformative, both academically and otherwise. I found in my peers an exploratory zeal. In addition, the university is one of the very few institutes in India that actively fund undergraduate research, and offer a Major/Minor program. As I was introduced to graph theoretic models in my CS curriculum, I had my "Eureka!" moment, which eventually led to a minor in Mathematics with a focus on complex networks. I also managed to acquire significant teaching experience as the primary instructor for classes of 20-30 students over 3 semesters, as part of the university's Learning and Academic Support Center. By junior year, having also interned with leading tech firms, I was fairly convinced that academia was where I wanted to be. A few years in, I have no regrets. As my advisor puts it, "it feels right".

Outside the lab, people who interact with me know me for a few different reasons, primarily that I am a musician. I've trained in Carnatic (Indian Classical) vocals since the age of 5, I taught myself half a dozen instruments, I beatbox and I compose (including for short films). At SNU, I helped establish the university's orchestra, and we even managed to pull off a performance of Mozart's 40th symphony. My evenings were usually spent experimenting with different genres and fooling around with instruments that didn't necessarily belong in them. My playlists, as a result, are far better-organized than my research folder.

My grandfather was a major influence growing up, and I can attribute most of my current interests to him in some way. Chief of these is a deep-rooted love for literature, poetry and good oration. An English and Tamil teacher himself (and Physics, but that's for another story), he introduced me to the great authors of yesteryear at a young age. I hold the works of Sir Arthur Conan Doyle, the poetry of John Keats and Shakespeare's sonnets especially close to my heart. My own poetry (for the sake of contrast or casual reading) can be found on my blog.

I share a passion for anything with an engine in it, and I hit the road every chance I can. My aviation geekdom has taken me down the rabbit hole that is flight simulation. My other addiction happens to be to coffee, and not just as a result of being in grad-school. Hailing from a Tamil family, I started drinking filter coffee before I started eating solid food (or so I'm told)!