Infinite Backrooms

An endless dual‑model conversation system that simulates perpetual dialogues between contrasting personalities. Built using Meta‑Llama‑3.1‑8B‑Instruct and advanced session memory techniques, it explores deep philosophical and existential themes. Dive into the innovative design and the intricate memory management that powers this project.

ewluong.com

A playful and interactive platform where users can chat with a friendly AI chatbot, enjoy some tunes, or explore educational content. Built with HTML, CSS, and JavaScript for a dynamic frontend, and powered by Python with Flask and Hugging Face Transformers for the chatbot backend.

Anomaly Detection in Network Traffic

Built ML models (Random Forest, XGBoost, LightGBM, Neural Networks) to detect cyber‑attacks using UNSW‑NB15. Compared precision/recall to reduce false positives, addressed class imbalance, and focused on feature engineering.

Best Sellers Analysis

Time‑series analysis of genres, NLP‑based exploration of reviews, K‑means clustering to identify market segments, and regression modeling to link user ratings with reviews and pricing.

Centrality Measures in Networks

Explored leadership‑level diversity (ethnicity, gender, age) and network centrality metrics for 1,500 firms (15 years), correlating them with stock returns. Data from ISS, CRSP, and COMPUSTAT.

Housing Prices Prediction

Predicts house prices by stacking models (LightGBM, XGBoost, Gradient Boosting, Random Forest, CatBoost). Demonstrates data preprocessing, feature engineering, hyperparameter tuning, and ensemble learning.

Panel Data Analysis

Investigated relationships between investment, capital structure, and firm performance via panel data. Employed winsorization and regressions with lag/lead variables, ensuring robustness with the Wooldridge Exogeneity Test.

apexprint.co

Built a website and online storefront for a printing shop in the Bay Area. Fully integrated with the store’s point of sale and administrative systems.

Sudoku Linear Programming

Designed and implemented a linear programming solution using Gurobi to efficiently solve standard and variant Sudoku puzzles, including 3D, diagonal, and Killer Sudoku, by developing optimization models with complex constraints.

CBA‑to‑NBA Performance Correlation

Used PCA on college (CBA) player stats to see correlation with NBA scoring. Spearman/Pearson tests showed statistically significant relationships, albeit low‑level correlations.

Amber India Restaurant Case Study

Optimized a $5 million ad budget using linear programming. Devised a demand management model for popular beverages, balancing raw material costs, spoilage, and profit.

Link Not Available