Darshan Thaker

Bachelor of Science, Computer Science

Bachelor of Science, Pure Mathematics

Turing Scholars, Class of 2018

UT Logo

Hi, I'm Darshan Thaker

A Computer Science enthusiast

Ever since I was a kid, I have been interested in Computer Science. I first started programming when I was in 3rd grade in Terrapin Logo. Since then, I have enjoyed expanding my interests in other programming languages including Java, C, Python, and a little bit of C++. In high school, I spent my summers interning at the awesome Open Networking Lab, where I got to learn about SDN (Software-Defined Networking).

After getting to college at the University of Texas at Austin, I began to explore my interests in Machine Learning and Artificial Intelligence, taking cool classes such as Computer Vision, Convex Optimization, and Deep Learning. I have been fortunate to work with some amazing teams through summer internships, such as the Center for Advanced Machine Learning at Symantec, the Holistic User Location Knowledge team at Google, and Facebook Search.

Outside of these interests, I competed in high school debate for 4 years, and I enjoy playing the piano to learn how to popular songs. I also love bouldering and have been doing a lot of that recently.

That’s pretty much me and my life right now, but as my life progresses, I wish to do research, go to graduate school, and make progress on some really fascinating problems in Machine Learning. My research interests are primarily in efficient large-scale Machine Learning at the intersection of Systems and ML, specifically with semi-supervised and transfer learning.

Relevant Coursework

Semester 1

  • CS 314 - Data Structures (Dr. Gordon Novak)

  • CS 311H - Discrete Math for Computer Science: Honors (Dr. Isil Dillig)

  • CC 307D - Intro to Roman Archaelogy (Dr. Denton Walthall)

  • UGS 302 - Economic IQ of the Media (Dr. Wayne Hickenbottom)

Semester 2

  • CS 429 - Computer Organization and Architecture (Dr. Robert Dickerson)

  • CS 178H - Intro to CS Research: Honors (Dr. Calvin Lin and Dr. Lorenzo Alvisi)

  • M 427L - Vector Calculus (Dr. Bart Goddard)

  • M 340L - Matrices and Matrix Calculations (Dr. Eric Korman)

  • M 362K - Probability 1 (Dr. Dave Rusin)

Semester 3

  • CS 439H - Principles of Operating Systems: Honors (Dr. Ahmed Gheith)

  • CS 377P - Programming for Performance (Dr. Sreepathi Pai)

  • CS 378 - Embedded Systems (Dr. Robert Oshana)

  • M 378K - Intro to Mathematical Statistics (Dr. Stephen Walker)

Semester 4

  • CS 378H - Distributed Computing: Honors (Dr. Lorenzo Alvisi)

  • CS 331H - Algorithms and Complexity: Honors (Dr. Eric Price)

  • GOV 312L - Introduction to US Foreign Policy (Dr. Robert Moser and Dr. Patrick Mcdonald)

  • SDS 325H - Honors Statistics (Dr. James G. Scott)

  • CS 109 - Competitive Programming (Jaime Rivera and Arnav Sastry)

Semester 5

  • CS 356 - Computer Networks (Dr. Simon Lam)

  • CS 345H - Programming Languages: Honors (Dr. Thomas Dillig)

  • M 365C - Real Analysis (Dr. Hector Lomeli)

  • EE 381K - Convex Optimization [Graduate] (Dr. Constantine Caramanis)

Semester 6

  • CS 343 - Artificial Intelligence (Dr. Scott Neikum)

  • CS 378H - Machine Learning/Computer Vision: Honors (Dr. Kristen Grauman)

  • M 361 - Theory of Functions of Complex Variables (Dr. Duncan Mccoy)

  • M 427J - Differential Equations with Linear Algebra (Dr. Bart Goddard)

Semester 7

  • CS 395T - Deep Learning Seminar [Graduate] (Dr. Philipp Krähenbühl)

  • CS 381V - Visual Recognition [Graduate] (Dr. Kristen Grauman)

  • CS 342- Neural Networks (Dr. Philipp Krähenbühl)

  • M 373K - Algebraic Structures I (Dr. Sean Keel)

A few of my projects


A parallel implementation in the MultiBoost library for the Adaptive Boosting (AdaBoost.MH) algorithm. This project was my final project in the "Programming for Performance" class.

Keywords: C++, Pthreads, Machine Learning, AdaBoost.MH, Parallel Boosting

Source code

A program to automatically generate music notes using a Markov chain trained on a sequence of notes. Uses music language ChucK to play generated notes.

Keywords: Java, ChucK, Discrete-time Markov Chain

Source code

A program to cluster your Gmail inbox into various topics based on the body text of the e-mails. Uses topic modeling algorithm LDA (Latent Dirichlet Allocation) to perform this clustering.

Keywords: Python, Latent Dirichlet Allocation

Source code

A shell script to aide in testing parallelized code. Compares the serialized and parallel version of code and outputs speedup gains from using different number of threads.

Keywords: Bash, Parallel Programming

Source code

Exploration of conflict graphs to parallelize stochastic subgradient descent in the context of Support Vector Machines (SVMs). This project was my final project in the "Convex Optimization" class.

Keywords: Python, Machine Learning, SVM, Stochastic Gradient Descent


Trained an ensembled deep convolutional neural network in Tensorflow using VGG19 and AlexNet architectures in order to predict what year a yearbook photo was taken. This project was part of "Deep Learning Seminar" course.

Keywords: Python, Deep Learning, Tensorflow, VGG19, AlexNet


Explored transferability of semi-supervised architectures by combining a ladder network with a progressive neural network. This project was my final project in my "Deep Learning Seminar" course.

Keywords: Python, Tensorflow, Semi-supervised Transfer learning, Ladder Network

Source code

Used Memory-Augmented Neural Networks to perform k-shot learning for video action recognition. This project was my final project in my "Visual Recognition" course.

Keywords: Python, Tensorflow, Memory-Augmented Neural Networks, One-shot learning

View full github

Research Interests

I am currently working towards my undergraduate honors thesis with Dr. James Scott from the Deparment of Statistics and Data Sciences at UT Austin. We are working on improving biomedical question answering systems, specifically tackling the BioASQ challenge .

My long-term research interests lie in the area of efficient large-scale machine learning, especially in semi-supervised and transfer learning.