Trained robust song embeddings with contrastive learning and data augmentation for recommendation systems.

Skills: Torch, MySQL, Contrastive Learning, Data Augmentation

Collaborators: Akshar Sarvesh

Trained graph convolutional networks and graph attention transformers to detect suspicious activity in transaction networks.

Skills: Graph Neural Networks, Torch, PyGeometric, Transformers, Feature Engineering

Collaborators: Alex Shan, Brent Ju

Trained a superhuman Liar's Dice agent with deep Q-learning and self-play, consistently beating benchmark agents with varied strategies.

Skills: Deep Q Networks (DQN), Torch, Reward Shaping, Belief Updating

Collaborators: Ryan Lian, Brian Xu

Conducted an experiment to investigate attribution bias for classical music between famous composers and artificial intelligence.

Skills: Artificial Intelligence, Experimental Design

Created novel matching algorithm and platform for music profiles, building full proof-of-concept with automatic matching.

Skills: React, Flask, MySQL, K-Means, Matching Algorithm

Collaborators: Professor Michael Bernstein (Advisor), Brent Ju

Developed a power set ensemble of fine-tuned BERT models for natural language processing tasks.

Skills: Torch, Multitask Learning, Cosine Similarity Fine-Tuning, Gradient Surgery

Collaborators: Jeanette Han, Kevin Song

Statistical analysis examining the key factors driving song virality and how they change over time.

Skills: R, Principal Component Analysis, Analysis of Variance

Created a movie chatbot that processes user input, performs sentiment analysis, and queries large language models for response generation..

Skills: Natural Language Processing, LLMs, Prompting, Python

Collaborators: Akshar Sarvesh, Miro Swisher

Implemented a dynamic memory allocator in C, managing the heap through efficient allocation, deallocation, and coalescing strategies.

Skills: C, Linked Lists, Memory Management, Valgrind

Data analysis across large COVID-19 datasets in BigQuery to identify the demographic and societal factors that increased COVID-19 case and death incidence rates.

Skills: BigQuery, SQL, Pandas, K-Nearest Neighbors, I/O Operations

Collaborators: Andrew Li

Applied social media curation principles to song recommendation in Spotify, validating collaborative filtering approach for playlist creation.

Skills: Python, APIs, Collaborative Filtering, Figma

Collaborators: Professor Michael Bernstein (Advisor)

App prototype for the resale of extra food to prevent food waste.

Skills: React, MongoDB, ReTool, Figma

Collaborators: Matt Vilaysack, Maria Wang, Peter Benitez

Establishing the case to extend war rights to non-state combatants in non-international armed conflicts in the Geneva Conventions.

Skills: International Relations, Ethics

Recreated and generalized the Wordle game (New York Times) for words between 1 and 10 letters.

Skills: Python, 2-D Arrays, String Manipulation

Assessing the macroeconomic and personal effects of adopting a 10% value-added tax in the United States.

Strategic analysis of pro-Biden super PACs and their ad spending data in the 2020 election.