Projects
Helping develop the UMN rocket team's proprietary WINGS software used for data visualization during and after rocket launches.
Being a full stack engineering project, I'm contributing by helping build a robust memory-safe Rust back-end that supports
live telemetry reception through a USB port and handles data processing, and a front-end written in TypeScript for data visualization
that features graph displays, indicator light displays, and more. Frameworks utilized include Tauri, SolidJS, and Tailwind CSS.
Helping to conduct end-to-end testing to ensure that the rest of the team can confidently visualize and interpret data in order
to fine-tune their projects.
Pre-processed over 30,000 images to help train a YOLOv8 object detection model to identify pieces of a user's outfit,
utilizing advancements in deep learning and computer vision. Used OpenCV for real-time outfit detection via webcam or
user-submitted photos and integrated OpenAI's GPT-4o-mini LLM to generate personalized outfit recommendations. Implemented
a k-means clustering algorithm for dominant color identification of each piece of the outfit in order to enhance recommendations.
Built a React front-end and a Python back-end with FastAPI to support the web application.
Application that enables users to interact with drones in a lifelike 3D model of the UMN campus. Users can schedule deliveries,
change the view of the front end, add additional humans or drones to the simulation, change the simulation speed, and show possible
routes for deliveries. Learned about and implemented various design and behavioral patterns to write code and implement a complex system,
adhering to SOLID principles. Learned various development processes. In the final month of working on the project, my group and I
implemented multiple extensions to the project utilizing the agile Scrum project management framework using Jira.
Data scraping project and web app that gives film recommendations for any Letterboxd user or a recommendation for two using Blend mode
by implementing a collaborative filtering based machine learning model utilizing SVD factorization for dimensionality reduction.
Users can include filters such as film popularity, film genre, and films in user's watchlist. Used asynchronous web
scraping techniques to collect ratings data from over 5,000 Letterboxd users, resulting in a dataset of approximately 13,000,000
ratings and 950,000 films.
Group research project in which several AI agents were created and tested to play the game Otrio, a variant of 3D tic-tac-toe,
specifically the three player version of the game. Methods utilized by the agents for playing Otrio include the Minimax and Max^n—
a variant of the Minimax algorithm generalized for more than two players— algorithms, Monte Carlo Tree Search, and a Deep Q-Network
(DQN), which is a reinforcement learning approach that utilizes deep neural networks to learn optimal policies. My primary contributions
to the project were the implementation of the DQN agent and writing the sections related to Q-learning, DQN, and Monte Carlo Tree
Search in the final paper.
A continuous project, I built this site to show off my work, school, and project experiences and also make it easier for
potential employers or collaborators to get to know me. I also viewed it as an opportunity to enhance my web development skills.
As of right now it only has a front-end, which is written in TypeScript and built using Next.js, MDX, and SASS for styling. For
ease-of-development I decided to use a component library, and I went with Once UI because of it's emphasis on easy styling
and accessibility.
