Justin Zeng

Selected Works

Throughout my career, I've worked on a variety of projects. Here are some that have shaped my path in software development and data science.

Mars Rover AI

Designed and implemented a robust graph-based search algorithm in Java to facilitate efficient pathfinding for a Mars rover navigating through diverse and challenging terrains. The project utilized various search algorithms including Breadth-First Search (BFS), Uniform Cost Search (UCS), and A* search to determine the optimal path from a starting point to a designated goal.

View Project

Stock Prediction with Deep Learning Sequential Models

This project assesses the effectiveness of various machine learning models -- linear, RNN, LSTM, and transformers (both encoder-only and decoder-only architectures) -- in forecasting stock prices. Leveraging historical stock data from Yahoo Finance, we focus on VOO, an S&P 500 Index ETF, to evaluate and compare model performance. Our experiment explores the strengths and limitations of each model in processing and contextualizing sequential data. Additionally, we offer actionable insights and discuss potential reasons behind our findings.

View Project

Multi-Layer Perceptron for New York Housing Market Classification

Implemented a multi-layer perceptron (MLP) from scratch to predict the number of bedrooms in a property based on various features such as broker title, property type, price, and more. Extensive hyperparameter tuning and performance evaluation were conducted to achieve high classification accuracy.

View Project

Interest Rate Predictor

Built a predictive model for interest rates using Linear Regression, Random Forest, and XGBoost. Enhanced predictive power through feature engineering and hyperparameter tuning. Achieved high accuracy in forecasts with cross-validation. Visualized predictions and trends using matplotlib and seaborn. Documented the workflow for reproducibility and transparency.

View Project

VGG16-LoRA-Brain-Tumor-Classifier

This research project explores the implementation of VGG16 neural network architecture enhanced with Low-Rank Adaptation (LoRA) for accurate brain tumor classification from MRI images. Check out our research paper to learn more about how we integrated LoRA's parameter-efficient fine-tuning approach with VGG16 to advance automated medical image analysis.

View Project