Collaborated on an optical flow estimation project, studying the DPFlow paper and codebase to understand its model architecture.

Reimplemented the core Bidirectional Dual Encoder module in PyTorch and integrated it into the existing codebase

Adapted and trained the modified model on a new dataset, gaining hands-on experience in model development and training with PyTorch

GitHub

Collaborated on implementing a Federated GAN based on the "Universal Aggregation Federated GAN" paper, applying federated learning techniques to a conditional GAN architecture.

Utilized PyTorch to preprocess and build the dataset, ensuring compatibility with the federated GAN training pipeline

Trained the federated GAN on a new dataset, evaluating performance using Inception Score (IS) and Fréchet Inception Distance (FID) metrics, with a result of 1.062 and 26.09 respectively

Collaborated with a team of 8 on an open-source Visual Studio extension, implementing Agile methodologies including daily stand-ups, weekly formal meetings, and sprint retrospectives to continuously improve workflow

Leveraged GitHub for version control, task management, and GitHub Action for CI pipeline implementation, ensuring efficient project coordination

Utilized Visual Studio API to interact with GitHub, facilitating data access and posting capabilities to enable code snippet sharing through GitHub Gist

Designed and implemented a user-friendly web view within Visual Studio Code using HTML, CSS, and JavaScript, enabling efficient code snippet management without the need for external applications or browsers

GitHub

Designed and developed a custom RDBMS in C++ as part of a two-person team, successfully managing 1,000+ lines of code and demonstrating strong collaboration skills

Implemented a SQL-like parser, enabling efficient data manipulation and retrieval

Handled low-level memory management for reading and writing operations

Leveraged Git for collaborative development, utilizing advanced features such as branching, pull requests, and code reviews to streamline workflow and ensure code quality

GitHub

Used solar data from NSRDB and electrical data from CAISO

Analyzed the demand energy and supply energy by solar irradiation in California from 2018 to 2020

Visualized various information by pandas and matplotlib

GitHub

Address the accessibility of smart speakers for individuals who are deaf or mute by applying gesture detection

Built a voice assistant utilizing Google API and YouTube API integrated with Python

Trained LSTM model capable of recognizing different gestures, on our own dataset, with accuracy 98%

Implemented Ubuntu as operating system on Raspberry pi and NVIDIA Jetson Nano

Awards:
Golden Medal, Seoul International Invention Fair;
First Place in AIoT Innovative Applications category;
International ICT Innovative Services, Merit Award, 2021 Student Engineering Papers Competition

IEEE Paper

Point 1

Point 2

Point 3