CS Projects

Developed an Instagram-like app using React and Flask, with dynamic post display, likes, and comments via RESTful APIs.

Implemented real-time interactions like infinite scrolling, double-click likes, and comment updates for a seamless user experience.

Developed LSTM and GRU models for stock price prediction, capturing temporal patterns in financial data.

Improved accuracy by integrating sentiment analysis with market data using advanced NLP techniques.

Incorporated stochastic elements into the models to account for market volatility, enhancing robustness and predictive reliability. 

Implemented a Variational Autoencoder using PyTorch to reconstruct images from the Flowers102 dataset, focusing on model architecture and parameter tuning.

Learned advanced techniques in deep learning including stochastic gradient descent, latent space optimization, and loss function customization to enhance image quality and model stability.

Developed a convolutional neural network (CNN) for classifying images of European landmarks, employing data augmentation and transfer learning to enhance model performance on limited training data.

Applied visualization techniques to interpret CNN decisions, gaining insights into model behavior and improving reliability through Grad-CAM visualizations and systematic architecture adjustments.

Implemented and optimized Support Vector Machines (SVMs) for sentiment analysis on Amazon Prime Video reviews, focusing on hyperparameter tuning, class imbalance, and bias detection.

Applied feature extraction, regularization, and evaluated machine learning models using metrics to address dataset biases and enhance prediction accuracy.

Explored adaptive tokenization in developing a dynamic algorithm to improve language models' adaptability in different textual domains

Introduced a creative approach to integrating vector semantics with adaptive tokenization.

Developed and implemented sorted, binary heap, and pairing heap priority queues using inheritance and dynamic polymorphism based on templated generic code.

Utilized branch and bound algorithm and Prim's Algorithm to solve the Traveling Salesman Problem (TSP) for a complete weighted graph, while exploring heuristic approaches for near-optimal solutions.

Created a relational database using structural query language (SQL).

Implemented essential functions including CREATE, INSERT, JOIN, REMOVE, PRINT, and INDEX GENERATION, with a focus on exploring binary search tree and hash tables.

Designed and implemented a real-time battle simulator, utilizing data structures such as priority queues and deque to efficiently process large volumes of incoming data.

Learned how to use stream-based algorithm, calculate running median, and identify “optimal” battles with given conditions. 

Created a program that is able to read, store, and solve a 3D maze.

Utilize Breadth-first search (BDS w/ queue), Depth first search (DFS w/ stack), map and coordinate list mode input and output, and handle command line arguments using getopt_long(). 

Wrote a program to automatically identify the subject of students’ posts on Piazza using natural language processing and machine learning techniques. 

Trained the data using the bag of words model and made prediction of tags using log-prior and log-likelihood probabilities. 

Handled binary trees with Containers ADTs, dynamic memory, The Big Three, iterators, and the map data structure.

Built a working web application accessible through a browser that directs the user to an office hour queue. 

Acquired skills to handle doubly linked-list with Containers ADTs, dynamic memory, The Big Three, and iterators. Learnt about simple Web 2.0 applications and HTTP. 

Developed a program that simulates a game of Euchre (a card game popular in Michigan) with a simple AI player.

•Gained experience in C++-style Object Oriented Programming (OOP) with classes and virtual functions that utilize Abstract Data Types in C++, derived classes, inheritance, and polymorphism. 

Built a content-aware image resizing program using a seam-carving algorithm by finding and removing “seams” in the image that pass through the least important pixels. 

Proficient in C-style pointers, arrays, structs, streams, IO, and Abstract Data Types in C. 

Wrote a program for RTMUS Lab to speed up cleaning and analyzing audio transcript from interviews and meetings. 

Utilized streams and classes to read in and classify important instances of each line in the transcript files. 

Designed a visual novel game with complex game logic utilizing the Ren'Py visual novel engine. The game contains tools to display thoughts, dialogues, menus, graphics, and music, and supports saving and loading of games. 

Used Python compound types like lists, tuples, sets, and Unicode objects; worked with Python-style objects like classes, functions, methods, and bound methods.

Fun fact: All the footages used in this project are pictures and videos taken during high school! Even the storylines are full of my memories!