Mural Design

Kroger Mural in Athens, GA

Digital Illustration

Floating By

Reading

Valley of the Shadow of Death

Kroger Mural in Athens, GA | Unused Draft

Traditional Art

Invisible Barrier - Self Portrait

The Places We Go

Frontend Design & Development

Data Analytics

Academic Research

Tableau

Microsoft Excel

RoadRisk Predictor: Leveraging Big Data & ML for Proactive Traffic Accident Prevention

Traffic accidents are a pervasive and devastating issue, with millions occurring annually across the U.S. While many factors are known, there's a critical need to move beyond general statistics and identify 

 that contribute to accidents. More importantly, the challenge was to develop a 

 that could assess the probability of an accident for a specific commute, empowering individuals to make safer travel decisions.

, a data-driven project that analyzed vast datasets of U.S. traffic accidents and corresponding weather conditions over multiple years. We leveraged 

 that, when integrated with real-time weather and location data via APIs, provides users with a percentage chance of encountering an accident on their chosen route.

Lead Frontend UI/UX & API Integration Specialist:

 Designed and implemented the user interface for the predictive model, allowing users to select locations on a map and view real-time risk assessments.

 Spearheaded the integration and management of both the 

 (for location input and route visualization) and a 

 Independently identified and acquired the extensive multi-year traffic accident and weather datasets for all 50 states, crucial for the project's foundation.

 Actively participated in overall project ideation and research, including reviewing academic papers on accident prediction and contributing to the understanding of key risk factors (e.g., black ice, motorcycle accidents, hotspot analysis).

Our project followed a robust data science and machine learning pipeline:

 Sourced and managed two massive, multi-year datasets: U.S. traffic accidents and comprehensive weather data across all 50 states.

 Performed extensive data cleaning and matching (likely using Python/Pandas, given the scale) to merge the disparate accident and weather datasets into a unified, analyzable format.

 Utilized Tableau to visually explore and identify correlations between various weather conditions (temperature, precipitation, visibility, wind) and accident frequency/severity, moving beyond obvious links.

 Collaborated with the team to develop a predictive model using 

 (e.g., Logistic Regression for baseline, Gradient Boosting or Random Forest for enhanced accuracy). The model was trained on the integrated historical data.

 Designed and implemented the frontend interface that consumes real-time weather data (via API key integration) and user-selected location (via Google Maps API) to feed into the trained ML model, providing an on-demand accident probability.

 Incorporated insights from academic research on accident prediction, black ice impact, and hotspot identification to inform model features and interpretation.

 Delivered a functional prototype capable of providing users with a personalized, real-time percentage chance of an accident on their commute, a novel application of traffic data.

 Uncovered nuanced correlations between weather patterns and accident occurrences, contributing to a deeper understanding of road safety risks.

 Successfully managed, cleaned, and integrated large, complex datasets, demonstrating proficiency in big data manipulation.

 Showcased strong capabilities in integrating external APIs (Google Maps, Weather) with a custom frontend and a machine learning backend.

 Developed a tool with the potential to empower individuals to make safer travel choices and potentially reduce accident rates.

This project provided extensive experience in 

big data processing, advanced data visualization (Tableau), machine learning model development, and complex API integration

. It underscored the power of combining diverse data sources and cutting-edge technologies to create practical, impactful solutions for real-world safety challenges. My role specifically honed my skills in frontend development and the critical art of API management.

Mobile App Development

UI/UX Design

Flutter

At the core of it, mind mapping is a visual notetaking strategy where ideas are written inside nodes and lines radiate out of central ideas to related ideas to show hierarchical connection.

Traditional handwritten pen-and-paper mind maps are one of the easiest and fastest ways to organize information. However, they have multiple issues:

They are flimsy, easily misplaced, and might get destroyed by spilling a cup of water or tearing the paper

If you wrote too much info to fit in the node, you gotta restart the mind map or erase the circle. Or just leave the text spilling out of the circle

For people writing as fast as their brains think, illegible handwriting ruins readability for future use

To share your mind map, you need to physically lend it or take a photo to share. It's one extra step but it adds up if this is your primary form of notetaking

Current digital mind maps solve all of these issues and offer great customizability that pen and paper can't offer. For example, multiple apps feature an infinite canvas, beautiful templates, and a plethora of design choices per each node, line, and text. However, these apps distract the user from mind map's whole purpose of being a rapid notetaking technique, and adjusting the basic nodes and lines are cumbersome and finicky. 

Collaborator Anisa Lastname and I created the digital SwiftThink app that addresses all of these concerns by integrating multi-touch and gesture. By using SwiftThink, making nodes and connecting lines take only a few seconds! Any concerns about aesthetics, and design decision paralysis are solved by the minimalist interface. The demo below proves just how fast and easy it is to use our app:

Canvas and ToolBar: Upon opening the app, the user will see a toolbar on the right and a blank canvas. The toolbar contains the necessary buttons for some key features. (See Figure 1)

The “N” button is for the Quick-Node Feature.

The “C” button is for the Quick-Connect Feature.

The Trash Icon is for the Quick-Delete feature. The Quick-Node, Quick-Connect, and Quick-Delete features will be explained later.

The 2-bar button is an expandable menu where the user can choose a rectangle or oval shape for the next node they put down. To make the user’s selection obvious, when the user presses on either button, a toast alert will slide down from the top of the screen to notify the user of which shape they selected. After a few seconds, the alert will slide back up to disappear, so that it will not distract from the user experience.

The Pencil Icon is a toggle-able edit feature that allows the user to enable or disable the nodes from being editable. When this button is green, the user can text in the node. When this button is red, the user cannot text in the node. This toggle feature was necessary because custom gestures are overridden by the default TextField gesture detectors, which would prevent the user from being able to do any features that need the tap or long-press gesture detection.

Custom Notes through drawing: This is an additional feature that we added to our mind-map application. The user can use the stylus to draw over the canvas, take custom notes, or add free-hand drawings to organize their thoughts and ideas in a customized manner.

Figure 2: Custom Notes Feature Demonstration

Quick Mind Map: The Quick Mind-map feature allows the user to quickly create a mind map. In our implementation, we used a long press gesture detector and multitap gesture such that when the user long presses on one node, that node becomes a “root” node. The user (while simultaneously long pressing the root node), taps on all other nodes they want it connected to, and they all are connected together.

Figure 3: Quick Mind Map Feature Demonstration

Quick-Node: The Quick-Node Feature allows the user to easily put down a node onto the canvas. In code, we used the long press gesture detector and multitap gesture detector such that the user must hold down the “N” button and simultaneously tap the screen in order to make a node appear at the tapped position.

Quick-Connect: The Quick-Connect Feature allows the user to easily connect multiple nodes with a quick swipe. In code, we used the long press gesture detector and pan gesture recognizer such that the user must hold down the “C” button and simultaneously swipe over the nodes the user wants to connect. The code will detect when the user’s pan touch has collided with a node. Let’s call this node A. Then when the user collides with node B in the same swipe, a line is created to connect the two nodes. The full overview of this gesture feature is demonstrated in Figure 2. 

Figure 4: Quick Connect Feature Demonstration

Holding the C button and swiping → node A detected → (still swiping) node B detected → create a connection line for A and B → (still swiping) node C detected → create a connection for B and C, and so on.

Quick-Delete: The Quick-Delete Feature allows the user to easily delete any node of their choosing. In the code, we used the long press gesture detector and multitap gesture detector such that the user must hold down the Trash Icon button and simultaneously tap on a node to delete it. Deleting a connected node will delete that node and the lines attached to that node.

Other Features - An additional feature that we implemented, which was not added in the proposal, was draggable nodes. The user can drag the created node anywhere on the screen, and if the node is connected to an existing mind map, the lines are dragged along the node.

3rd Party Libraries / Packages / References

https://pub.dev/packages/overlay_toast_message

https://pub.dev/packages/one_dollar_unistroke_recognizer

Co-led the technical development of the application using Dart and Flutter.

Spearheaded the integration and design of multi-touch interactions as the core innovation for rapid mind map creation.

Contributed significantly to the overall UI/UX design, focusing on simplicity, speed, and intuitive user flows.

 During fast-paced demonstrations, SwiftThink proved significantly faster at creating and organizing mind maps compared to regular single-touch digital applications.

 Enabled users to capture ideas at the speed of their brain, minimizing the loss of fleeting thoughts due to slow interface interactions.

 Eliminated handwriting barriers, making mind maps consistently legible and easily shareable for collaboration or review.

 Successfully addressed cumbersome workflows and choice fatigue prevalent in other mind mapping tools, offering a truly intuitive experience.

 Successfully learned and applied Dart and Flutter, showcasing the ability to master new technologies and implement complex multi-touch interactions.

Full Stack Development

Spring Boot

Java

ReactJS

OpenAI API

Figma

GSAP

Typescript

NextJS

TailwindCSS

Sanity IO

Rive

Adobe Illustrator

Frontend Web Design

Wix Website Builder

Visual Artist

Photoshop

The Georgia Tech Computer Science capstone course is a year-long software design project that aims to solve real world problems. Projects are led by clients and professionals within the GT community. For our team of 6 students, Professor and Korean Language Director Yongtaek Kim asked us to find en engaging, gamified edutech solution to teaching his KOR 1001 class.

Students dread language learning if it feels like a chore, and this is often the case when class assignments consist solely of reading a textbook and completing worksheets. Lingual and cultural interactions are a tried and true way of engaging students in language learning, but this is difficult for professors to implement without actually taking their students abroad. The goal of our project is to create a fun and engaging application that helps students learn both the Korean language and culture. Current language learning applications like DuoLingo exist, but they do not teach the cultural aspect of languages.

Dr Kim has been hoping to solve the problem of demotivated students for years now, and a solution idea presented itself as the Korean language and culture has rapidly grown in popularity. The solution takes inspiration from the vastly popular Netflix show called Squid Game. Squid Game is a Korean drama in which characters are faced with deadly stakes while competing in Korean children's games in hopes of securing a huge prize. Given that Squid Game is now the most-watched Netflix show of all time, Dr. Kim had the idea to utilize its momentum and create a fun beginner's Korean game application that consists of language learning minigames created based on the games from the show, such as red light green light and tug of war.

Our product vision is to create an application that combines the best of all worlds. That is, we want our application to include study features similar to DuoLingo, but we also want it to be tailored specifically to KOR 1001 and include minigames that will make using the app fun. In addition to the learning minigames, the application will include flashcards for each unit of Korean 1001, as we hope to make our app a one-stop-shop for all the studying and learning needs of its users. Overall, we hope the auditory, verbal, and visual features of our app will provide an engaging experience and make it so that studying for Korean class feels more like a hobby than a chore.

When we first started this Junior Design Project, we had to research different aspects of this project to create a game app that met Professor Kim’s goals.

We decided to use a Google form poll to complete our primary research because it is a quick and convenient way to gather information, and because Professor Kim has access to the Korean course Emailing list, we were able to send it out to our primary audience: Korean students.

Based on our user research and surveys, we came to the conclusion that there is broad support among GT Korean students for a game-based language learning app, with an overwhelming majority saying that this app would be a needed/wanted study tool. Also, through our survey, we also found that repetition is a heavily requested feature. Audio recordings of vocabulary and questions were also requested multiple times. In addition, flash cards to practice vocabulary were explicitly mentioned in the responses to the survey. All of these responses show us that there are a few core features that users care about.

Additionally, we read 2 articles as our secondary research to better understand the development of a Korean language app. We wanted to research about how other language apps in the past were developed to find out what the most common trends in language-learning software are. We knew that it was extremely important to be up to date on recent software and applications, so that we can learn from past projects and better implement our own app.

Through our secondary research, we became up to date on modern language apps and what users want most. Trends show that PC apps are currently the most popular. However, with the increasing availability of smartphones, mobile apps are becoming more standard. The most frequent users of these apps are university students, with most users reporting positive outcomes and language learning growth. By looking at other applications, such as the mobile game “Infinite Korean”, and by reading user reviews, we also found that people prefer auditory repetition as a language learning feature. Additionally, users report that having the choice of toggling romanization on or off as a feature was preferable for language apps.

To accomplish this, our UI has an easy interface to understand and use as we will show in our live demo at the end. Each unit of the KOR 1001 curriculum is clearly available to the user, and the user can decide to study grammar and vocabulary through their game, or study vocabulary through flashcards as we will show in our demo later.

For our learning game we decided to create our own interpretation of the red light green light game from Squid Games. It was the most iconic game of the show and it has easy enough rules to follow.

In our version of the game, the robot doll will say a Korean word or phrase and the user must select the correct word or phrase that has been said. If they get it right, they will get a point and their goal is to get 10 points. If they get it wrong, they will lose a life, and they have 3 lives before they lose the game. They can also lose the game if they do not finish within the time limit.

This will hopefully create an engaging way for users to learn through an auditory component as well as through repeated use of words and phrases as the students we surveyed had requested.

 Collaborated on designing intuitive navigation and adhering to standard UI/UX principles for button placement and overall site flow.

 Contributed to the front-end development of the React-based website.

 Responsible for all visual assets, including Squid Game-themed character drawings, homepage artwork, assassin wallpapers, and the website logo. Utilized a vibrant color palette directly inspired by the series to ensure visual appeal and engagement.

, focusing on end-to-end development from user research to iterative prototyping and milestone delivery.

 Collaborated with the professor to understand pain points in language learning and brainstormed gamified solutions, drawing inspiration from the popularity of Squid Game.

 Designed and produced all custom artwork in Adobe Photoshop and Procreate, ensuring a cohesive and visually striking aesthetic that aligned with the Squid Game theme. This included adaptive designs for wallpapers and homepage art to ensure natural scaling across different screen sizes.

 where correct answers propelled the user's team forward, adding a dynamic, competitive element to vocabulary recall.

 Implemented Quizlet-like flashcards and quizzes, centralizing vocabulary and audio resources for easy access and review.

 Addressed challenges related to image sizing and responsiveness to ensure all custom artwork scaled naturally across various viewports.

 Created a unique, visually captivating learning environment that significantly increased the potential for student engagement in Korean language studies.

 Provided a centralized, user-friendly platform for all Korean 1001 learning materials, reducing reliance on disparate external tools.

 Introduced gamified elements designed to foster healthy competition and motivate consistent vocabulary memorization.

 Successfully delivered a functional web application within a real-world academic context, demonstrating proficiency in team-based software development and creative problem-solving.

Since the creation of this project by our team, other teams have since worked on the website, adding new designs and features. See the current live website here: https://korean-game.modlangs.gatech.edu/

Rina Yoo

Frontend Designer & DeveloperAssociate Lead Mural DesignerFreelance ArtistCat Lover

U.S. Traffic Accident ML Predictor

Using weather and traffic accident data analytics + machine learning to predict the likeliness of a traffic accident

SwiftThink App

Rapid Multi-Touch Gesture-Based Mind Mapping

Acadess

AI Burnout Prevention Journal for Students

CS Portfolio

Korean Language Game

This Resume Website!

Alaska AJU Tour by Hyung