Kyoto 2021
Walk through the forest to the Kifune Shrine
(Kyoto, Japan)
DREAM Graduate Student Researcher
Under the guidance of Dr. AlQuaraishi
Researched deep learning neural networks predicting 3D RNA structure in collaboration with Harvard Medical School
Amazon Software Development Intern
My first tech job!
I joined Amazon Ads as an SDE intern in June 2022. Having zero experience in neither tech nor the corporate world, I was extremely excited yet nervous at the same time. However, my manager and team were all extremely helpful and supportive throughout the process. I asked so many questions, some very basic, and thanks to everyone, I don't think I've ever learned so much in such a short period of time.
Artificially Intelligent Snake?
Deep Reinforcement Learning to Play Snaket
AI learns how to play snake using reinforcement learning given only the basic rules and possible actions. In reinforcement learning, the AI agent learns optimal policies from samples or games outcomes. The agent then uses the acquired knowledge to make better decisions in the subsequent games. This model uses a 2-layer neural network to predict Q-values (utility) from the current state, which are used as target values in the Q-learning strategy. Q-learning is an off-policy approach, which is considered to be rather general and useful to learn about arbitrary policy.
Classify Handwritten Digits
Use Machine Learning to Read Numbers
Trained on the MNIST dataset, this Convolutional Neural Network (CNN) program allows users to upload their own handwritten numbers and outputs the value. The network was implemented using PyTorch and uses libraries such as NumPy, pandas, and Matplotlib. You can test your own handwritten images. Just take a square image of the number you wrote, and add the .jpg file to the images directory. tester.py will automatically run and check it.
Train Accuracy: 99.7%, Test Accuracy: 99.1%, Custom Test: 9/10
Cryptopoly
HoyaHacks 2022 Build A Game Track Winner
Developed a cryptocurrency-themed monopoly web app game using HTML/CSS/JavaScript for 2 players. Property values fluctuate based on real-time cryptocurrency price changes, which are web scraped using Python. My role was designing the game logics, including the game board, player interactions, property states, etc. Even though this was our first hackathon, we fortunately won the Build A Game Track!
Columbia University
CS@CU MS Bridge Program
In Fall 2021, I started the CS@CU MS Brige Program at Columbia University, Department of Computer Science. This program is designed for MS CS students who do not have a CS background. Since I had taken some math and CS courses in undergrad, I took Data Structures, Discrete Mathematics, and Advanced Programming. In addition, I am also taking grad level courses: Artificial Intelligence and Introduction to Databases.
Binary Neural Network
My First Software and AI/ML Experience
In Fall 2020, I starting researching convolutional neural networks at The University of Tokyo (Tokyo, Japan), while also getting my first experience in CUDA, UNIX, and TensorFlow/PyTorch. I researched Binary Neural Networks (BNN) and looked at how different network architectures can affect the efficiency and accuracy.
I developed a BNN that saves tensors from the shallow part of the network, then applies a classifier on each saved tensor, and merges the different predictions. This idea evolved from ResNet and skip connections, believeing that there are benefits in keeping information learned in the shallower part of the network. Ultimately, I was able to decrease parameters by 63% and 27% for CIFAR-10 and CIFAR-100, respectively, and yielded accuracies of 90.72% and 67.65%. Despite using considerably less computations (about a third to a half the size), my model was able to outperform similar BNN and IR-Net.
Eye Stuff
Biomedical Research
Upon graduating from UT Austin, I worked as a biomedical engineer at the Petroll Lab at UT Southwestern Medical Center (Dallas, TX). I researched the correlation of focal adhesion and migration patterns of keratocytes in varying ECM topographies and growth factors. Furthermore, I also analyzed post-operational rabbit eyes using in vivo confocal microscopy to track surgical recovery.
The University of Texas at Austin
B.S. Biomedical Engineering, Class of 2019
I enrolled in UT Austin, Biomedical Engineering in Fall of 2015. During my time at UT, I was an undergraduate researcher, in which I received a $1000 grant from the Undergraduate Research Fellowship to study dermal heat conductivity and simulate the process using CAD and 3D printing. I was also involved in UT SEEK, an after school program promoting awareness of higher education and STEM to underprivileged kids in the greater Austin area. We paired with a student and met every week to build a STEM project, like marble roller coasters to teach kinetic/potential energy and ice cream to physically experience thermodynamics. In my final year, I was the Vice President of Corporate Affairs and successfully raised over $13,000, almost three times our previous annual budget.
Thanks Neil for the photo :)
Travel, Hobbies, Etc.
Cherry Blossoms 2021
Cherry Blossoms in Shinjuku Gyoen National Garden
(Tokyo, Japan)
Ski 2021
Ski trip to Maiko Snow Resort
(Niigata, Japan)
El Escorial 2019
Historical residence of the King of Spain ft. Dad
(Madrid, Spain)
Malaga 2019
Strolling through Malaga
(Malaga, Spain)
Zion National Park 2019
Views overlooking Zion National Park
(Utah, USA)
Brooklyn Bridge 2018
Looking up to Brooklyn Bridge
(New York, USA)