My Story Disguised as a Research Paper
Abstract
At the end of 2021, after having a beer, Mingi Kwon had an amusing thought: “What if I wrote my CV like a research paper?” Right there, he started drafting a CV titled “A Consideration of a CV in the Format of a Research Paper.” A lot of time has passed since then, and after looking at how others do it, he created a new, more formal and businesslike CV page. Still, he decided to reserve a corner for a space where he could write extensively about himself. This document is written in the format of a research paper to showcase ‘me’ more than any other CVs. It’s a concept. This document is helpful for those curious about who Mingi Kwon is. All the facts are based on truth, and it is a state-of-the-art document about Mingi Kwon. (Most of it is copied from past CVs. It’s quite embarrassing. What was I thinking back then? But I must leave it as it is—it’s fun, after all. Cheers!)
If you want to see my story, refer to the Introduction. If you are curious about my skills, refer to the Method section. For my career experience, see the Experience section. My published papers will be added to the References.
Introduction
Describing oneself to others is very difficult. Even long-time friends often don’t fully understand who I am, so it’s much harder to tell someone who has never met me. If it’s just a few lines of text, conveying my true self becomes even more challenging. Through this document, I aim to convey who I am as fully as possible. It’s not something for submission to a company, nor does it follow a specific format. (Though I’m using a research paper format, it’s simply my freely chosen style.) Thus, to the unspecified masses reading this, I want to convey my capabilities and who I am through my story.
I’ve been doing something similar to coding since I was young. It started with impure motives: I first started programming to bypass parental controls on the computer. This evolved into finding ways to use the internet secretly in the school’s computer lab. I even remotely shut down friends’ computers and created programs that displayed countless warning messages, once even shutting down all the computers in my middle school. However, I did all this without realizing it was coding, and I never considered becoming a developer. Coding was just a mischievous boy’s toy. My lack of interest in boring things led me to neglect my studies, resulting in my admission to Kangwon National University’s Department of Electronic Engineering, one of the lowest-ranked schools compared to my peers who went to prestigious universities.
In the second semester of my freshman year, I took a C programming course. It was the first time I formally learned coding, yet it was something I already knew. I found it interesting and enjoyable. That’s when I started to consider becoming a developer. To test this direction, I sought out programming jobs before enlisting in the military and managed to work as a freelancer for about three months thanks to an introduction from an acquaintance. This strengthened my interest in becoming a developer. One memorable anecdote: the company president asked me to write a specific manual, giving me two weeks. The task involved a lot of copying and repetitive work. Instead of doing it manually, I spent a week coding and completed the entire manual with a single press of the Enter key. The president, a really nice person, told me to take a week off, as he had expected it to take two weeks.
My interest in artificial intelligence started during my military service. While serving in the 7th Infantry Division, I had a lot of personal time as a corporal. Not watching TV, I opted to take online courses. Through MOOC, edX, and edwith, I took AI courses for no particular reason. Back then, AI wasn’t as widely known, so I mostly took foreign courses and even a course offered by Korea University. I was fortunate to receive the College of Engineering Dean’s Award for being the most diligent student in that course (this was back in 2016).
People often told me I was smart and, but I just had a good head on my shoulders. I never did what I didn’t want to do. I liked math but not English, leading to a high school report card with a grade 2 in math and grade 8 in English. Though I didn’t study hard in college, I earned high grades. My final GPA was 4.2/4.5, which might suggest a diligent college life. In reality, I simply enjoyed attending interesting classes. My major courses were fun, and my grades reflected my enjoyment and aptitude for the subject. During my spare time, I studied AI bit by bit. Even after my military service, I continued taking AI-related courses. I didn’t really study; I just watched interesting videos. Watching science and math videos on YouTube was a hobby, so these lectures were just another form of entertainment.
My desire to apply what I knew led to a grand scale for my senior year graduation project. Inspired by a conversation with a pianist friend who said it took about two weeks to find his own fingering for each new sheet music, I created a system that recommends fingering for new sheet music. Using videos, the system recognizes the piano keys and the fingers of the person playing. It identifies which keys are pressed when and which fingers are on the keys using an electronic piano. It records how long each key is pressed and which fingers are used, creating a dataset. Using the rhythm, notes, and previous finger numbers, it trains a model to predict the next finger number. When given new sheet music, it recommends fingering based on the rhythm and notes. Although there were many things to improve in retrospect, the project was completed and worked. It won the grand prize among all graduation projects, earning me the President’s Award from Kangwon National University.
After graduation, I immediately sought employment but found that AI-related fields required at least a master’s degree, prompting me to pursue graduate studies. Currently, I am enrolled in the integrated master’s and doctoral program at Yonsei University.
I have a personality that cannot tolerate inconvenience. Throughout my life, I’ve created various programs out of necessity. Before COVID-19, I attended church every week, where attendance was manually recorded. Our department alone had about 500 members, and watching them record everything by hand and then transfer it to Excel seemed inefficient. I created a simple website where team leaders could check attendance, automatically recording it in Google Sheets. This system was used for over two years, aiding in team member organization and other tasks.
Upon entering graduate school, the first thing I did was, unsurprisingly, read papers. AI papers are most frequently downloaded from arxiv.org, where the file names are just numbers. To avoid the inconvenience of renaming files, I created a Chrome extension that automatically names files with the paper title, author, and year. The user base has grown to nearly 1,000 users.
I often needed to check the status of GPUs used in the lab and find available ones for running tasks. To streamline this, I created a program that records the status and users of various servers’ GPUs in a specific Google Sheet. This allowed us to check the current status at a glance, and it is still used with great satisfaction.
I also created a program to automate the creation of manuals during my freelance period and automated daily repetitive tasks with a single click in the military. (This was later removed during an inspection since unauthorized programs can’t be used in the military.)
So, let’s summarize who I am. I excel at interesting tasks and dislike inconvenience. I proactively and creatively solve problems and often don’t engage in tasks that aren’t enjoyable.
I consider myself someone who translates thoughts into actions. I’m a fan of the YouTube channel Veritasium, which presents science and math topics with high-quality videos. I thought it would be great if such videos were available in Korean, so I dubbed Veritasium videos in Korean and sent them to the original channel. After some discussions, I started running the Korean channel, which now has around 200,000 subscribers.
Translating foreign videos into Korean made me understand what’s important and needed in this process. I thought that using AI, which I research, could make video translation more accurate. Thus, while pursuing graduate research, I’ve developed and refined various technologies for translating videos as a side project. Reflecting on my journey of writing numerous papers, running a YouTube channel, and continuously working on side projects, I feel quite accomplished. This effort culminated in the founding of GivernyAI. GivernyAI is a collection of AI technologies that translate languages, translate text on screens, and even make people on screen speak in the target language. The technologies include STT, LLM, TTS, Video Inpainting, Talking Head, and Audio Separation, each of which could be a startup on its own. I have consistently developed and improved these technologies.
I want to deliver “knowledge” to people fairly. Just as I struggled with English and found studying difficult, I want to ensure everyone can access high-quality videos in their native language.
Related Work
In a traditional research paper, the Related Work section introduces other papers in the same field. However, for a CV, this is not possible, so I will describe works related to the introduction. The introduction focused on coding and artificial intelligence, but many stories were left out.
Firstly, I won a gold medal at an international invention competition. I participated in a hackathon where I created a device that uses lasers to indicate escape routes in case of fire. Using the property that lasers are more visible in fog, the device intuitively guided directions through a moving laser towards the emergency exit. This was also filed as an international patent. I was selected as one of the eight students from Kangwon National University to study in the United States. I attended an 8-week leadership camp at Bloomfield College in New Jersey and took programming classes from a developer at Moody’s in New York. I have 4 weeks of internship experience at a database-related company and hold a certificate in public data utilization.
As a personal hobby, I dub and upload videos from the overseas YouTuber Veritasium into Korean. If you are curious about the YouTube channel, click the YouTube button at the bottom of the page. (Subscribe and like?)
I've written a lot of papers. It seems I have lived an incredibly hardworking life.
(Co-First Author) "Understanding the Latent Space of Diffusion Models through the Lens of Riemannian Geometry" [NeurIPS 2023]
(Co-First Author) "Training-free Style Transfer Emerges from h-space in Diffusion Models" [WACV 2024]
(First Author) "Diffusion Models already have a Semantic Latent Space" [ICLR 2023 spotlight]
(Second Author) "FurryGAN: High Quality Foreground-aware Image Synthesis" [ECCV 2022]
(Co-First Author) "Unsupervised Discovery of Semantic Latent Directions in Diffusion Models" [Just arxving]
(Co-First Author) "Attribute-Based Interpretable Evaluation Metrics for Generative Models" [ICML 2024]
(First Author) "Coming Soon: Text-to-Video Paper"
(Co-Author) "Coming Soon: Motion Customization Video Diffusion Model Paper"
(Co-Author) "Plug and Play Diffusion Distillation Paper" [CVPR 2024]
There are a lot of co-first author papers. In fact, all those papers were written since the latter half of 2023. The two years of my master's program seemed to be a learning phase, and afterward, I wrote many co-first author papers during the period when I had many things I wanted to do but not enough hands. I extend infinite gratitude to the co-authors who wrote the papers with me.
In the meantime, I won the gold award at the Samsung HumanTech Competition. This is a big award. I was also invited as a speaker at the Seoul AI hosted by the Seoul Metropolitan Government, but due to not being able to answer the phone during the conference period, I was demoted to a general presenter. They say the Mayor of Seoul, Oh Se-hoon, was there, but I didn't see him. Haha.
Currently, I am also engaged in a startup. I hope GivernyAI / Cinelingo does well.
Method
I have worked with various programming languages. I have experience with C/C++, Java, Python, JavaScript, HTML, CSS, PHP, R, SQL, ASP, and VB. In web development, I have used Spring and Angular, and I can make them work with databases. I have used C/C++ for embedded system development. With my knowledge of electronic engineering, I can design circuits and code data communication for various sensors. I have also developed Android applications and done FPGA coding.
However, recently, I have only been coding related to artificial intelligence using PyTorch, so I seem to have forgotten other skills... I use Linux freely and can handle various GPU settings without any problem. I know how to use Docker and Singularity and quickly adapt to different environments.
It might sound boastful, but I think I am good at coding. More precisely, I can quickly read and follow others' code.
Experience
I have been living the life of a graduate student. I have learned and acquired a lot under the excellent guidance of Professor Young-Jung Oh. This includes not only knowledge and research methods but also character and how to handle people.
I have six months of internship experience at Adobe. I spent six months in Silicon Valley, USA. I was the only Adobe Research intern whose in-person period was extended by three months, making it a six-month tenure.
I am the co-founder and CTO of GivernyAI.
Conclusion and discussion
Who am I? Reading this, I might seem like a person who boasts endlessly. But the message I want to convey is simple. To me, coding is a tool I use to solve various problems. I am someone who pursues fun and loves science and math. I tend not to engage in uninteresting tasks, but when I find something intriguing, I tackle it with a creative perspective. I am also a dreamer, living life with grand dreams. Additionally, I really enjoy having fun. I can play various sports like billiards, bowling, table tennis, basketball, and soccer to a decent level, and I enjoy karaoke, board game rooms, comic book cafes, PC cafes, etc. I even have experience participating in and winning a Hearthstone game tournament. I enjoy watching YouTube and like drinking. However, I prefer small, quiet gatherings rather than noisy drinking parties.
I spend quite a lot of time on these leisure activities. Anyone, in any format, is welcome to contact me for a cup of coffee, a beer, a cocktail, whiskey, wine, or makgeolli. (I don't like soju.)
Appendix
Education
Banwon Elementary School
Kyungwon Middle School
Hyundai High School
Kangwon National University
Yonsei University Graduate School
Military Service
Full-term discharge from the 7th Infantry Division, 8th Regiment, Reconnaissance Company (inside the DMZ), ROK Army
Affiliation
Yonsei University Graduate School of AI, GivernyAI/Cinelingo CTO