Projects

Awarded Top Project in Robust Algorithms for Machine Learning for developing novel algorithms to estimate edge probability in adversarially perturbed Erdős-Rényi graphs.

Introduced the $O(q,\varepsilon)$-adversarial model, and designed and implemented three novel robust algorithms

Conducted theoretical and empirical analysis to establish error bounds and runtime guarantees

Participated in the Citadel Invitational Datathon 2024

Conducted a predictive analysis on fast food access, socioeconomic factors, and obesity across U.S. states and counties

Developed and evaluated predictive models (e.g., Linear Regression, Random Forest, LSTM), identifying key obesity drivers using SHAP and LIME analysis

Built a friend-matching website prioritizing matches based on personality using feedback from over 20 users

Designed a matching algorithm and wrote a Blackjack mini-game with function to describe risk-adverseness of user

Engineered with a team of 5 using React and Java

Designed and trained a novel lightweight CNN, LYTNet, to detect pedestrian traffic lights and position of zebra crossings

Achieved 30% lower error rate compared to existing methods (96% accuracy; angle error rate of 6.15), while running at similar inference speeds (16.34 frames/sec)

Coauthored 2 papers; keynote presentation at CAIP 2019 and poster presentation at ICCV workshop 2020

Developed a WeChat Mini Applet to digitalize and centralize information for the SAS carnival

Resulted in ~2200 users (~1000 concurrent) and processed transactions worth over $35k of revenue for school

Cut down post-analysis time including time spent counting tickets by 90%

Hosted school assemblies for announcing logistics and advertisement as the President of Executive Student Council

Implemented a semantic segmentation model in PyTorch for accurate detection of traffic lanes

Fit the lanes to the nearest cubic polynomials using regression analysis