N Kim portfolio

About Me・自己紹介

I have a Bachelor of Computer Science from Thompson Rivers University in Canada, Bachelor of Commerce　from UBC in Canada, MSCS from Georgia Tech and speak fluent Japanese, some Korean and a little bit of Chinese. At OMSCS I completed Software Analysis and Test so have a good background in static and dynamic analysis and testing including instrumenting LLVM IR and using symbolic execution engines such as KLEE. Other courses I completed in the past year include Game AI (Unity) where I used fuzzy logic and neural networks (ML Agents, PPO algorithm) to train a race car to drive on its own, Information Security where I learned how to mitigate OWASP top ten threats, binary exploitation, RSA exploits and other security issues. I have 6 or 7 web apps deployed and some experience with Kaggle competititons. I also have other projects in computational photography, AI for Robotics and machine learning. My other interests in computing include GPU computing using Cuda and theoretical computer science, bioinformatics, robotics algorithms and functional programming. Nowadays however I am developing pipelines for the automation of subtitle transcription and translation, using AI to clone and generate voices for dubbing. I am also learning Unreal 5 for film making so I can get into virtual production as a career. In my free time, I enjoy cycling, working out, listening to music and watching movies.

Experience・経験

Freelance

Developer/Translator/English Teacher/Tennis Coach/Personal Trainer

Since June 2024, I am a full time translator and developer for a small movie production company in Seoul, Korea. This is quite a varied background which I feel helps me as a developer since I bring many different perspectives to solving problems. For instance my teaching background helps me write code that is highly readable since I am used to writing things down in a way that my students could follow. This also makes it more modular and reusable. As a translator, I developed strong research skills so anytime I have run into problems while coding, I have always been able to Google my way to working solutions. I also have a good grasp on linguistics which helps me when building neural networks for NLP (natural language processing). As a tennis coach and personal trainer, I know the value of chipping away at problems bit-by-bit and making incremental improvements to myself and to code I write which accumulates over time into positive results.

Education・学歴

Georgia Institute of Technology・ジョージア工科大学

Sept 2019 - July 2023

Masters of Science in Computer Science

コンピューターサイエンス学部大学院

Overall GPA >= 3.75 成績平均　>=　3.75

Thompson Rivers University・トンプソンリバーズ大学

Jan 2017 - Dec 2020

Bachelor of Science in Computer Science

コンピューターサイエンス学士号

Overall GPA >= 3.43 成績平均　>=　3.43

University of British Columbia・ブリティッシュコロンビア大学

Sept 1994 - June 1999

Bachelor of Commerce in Finance

商学部金融学科学士号

Overall GPA >= 3.3 成績平均　>=　3.3

Projects・主なプロジェクト

AI voice cloning and generation for English dubbing of anime

Python for subtitle transcription and translation, video processing, speaker diarization, voice cloning and generation. Source video audio was Japanese and voices were cloned from the Japanese source audio.

Link to video clip

OpenAI ChatGPT subtitle translator app

React frontend, Python FastAPI backend that calls the OpenAI api endpoint to translate srt files segmented, with custom prompts for each series and custom parameters.

Link to app YouTube

Malware classification Streamlit App

Deployed on Streamlit I classify malware using Python Scikit-learn comparing Random Forest, Gradient Boosting and Linear Regression.

Link to app

Static Plant Based Food Website

Made with HTML5, CSS, Bootstrap and a tiny bit of JavaScript to make the case for plant-based diets. The font is difficult to read but I just treated this as an experiment to try out different fonts downloaded, layouts such as parallax, etc.

ビーガン（粗食主義）の食生活がどう温暖化対策に貢献できるか、そしてビーガンの健康効果を説明しています。 Link to ecoWarriors static site

Bioactive Food Ingredients

Made with PHP, JavaScript, CSS, Bootstrap, HTML5, MySQL and deployed to Heroku (was deleted), this app focuses on disease prevention through diet based on data taken from a Japanese university's databases using Python BeautifulSoup and translated using Google Translate API (it was GCP or Azure) services accessed via a PyPi package.

Link to bioactive peptides app repo

Bioinformatics sequence alignment app

Python Plotly Dash bioinformatics web app with CI/CD pipeline (CircleCI and pythonanywhere) and browser testing with Selenium was made for my final project in an intro to bioinformatics class.

Link to bioinformatics app

Canadian food database app

This was made with Python, Plotly Dash and MySQL and MongoDB. However, the app in development is shown operating in the following videos. Also, the repo is listed on my resume.

video part 1 video part 2

C++ Qt Calculator App

Made with Qt GUI and C++, this calculator uses the Exprtk Math Engine under the hood. My calculator is fairly basic but can be expanded to do more. The repo is listed in my resume.

Link to C++ Qt Calculator repo

Ninja Survival Game made with Unity

A ninja survival game made with Unity 3D game engine. Please see below.

Robotics AI Projects・ロボット工学のAIプロジェクト

Kalman Filter カルマンフィルタ

The video shows a simulation of a spaceship navigating an asteroid field using Kalman filters to detect asteroids and predict possible collisions so it can get from the start at the bottom to the goal at the top.

Spaceship navigates asteroid field video Project Description

AI Racecar made with Unity

Using fuzzy logic to autonomously race a truck in Unity. Also used Unity ML-Agents PPO algorithm (reinforcement learning) to train the race car.

fuzzy logic race car neural network race car

Computer Vision, Perception Projects・コンピュータービジョンのプロジェクト

Face Landmark Detection

This was more about tuning open source face landmark detectors to work with my dataset but as I have written a dog breed detector in Tensorflow and Keras, I am confident I can replicate this quite easily.

Link to face landmarks video

CNN for Defect Detection in Windshields

I passed a test for a German Manufacturing Execution System company's test for machine learning engineers. I used transfer learning with the VGG16 backbone and data augmentation to find the cracks in the car windshields.

Link to video part 1 Link to video part 2 link to code repo

Particle filter on tennis player

I wrote a particle filter that uses Gaussian probabilities to localize objects. There is a demo video for the Mars Glider simulation as well as for using a particle filter to localize tennis players on broadcast TV tennis. I added another link to a tennis video where I used Tracknet to track the tennis ball and Openpose to track the players' poses.

Link to particle fitler video Openpose and Tracknet in tennis video

Seam carving images around key points

The midterm project, seam-carving by Avidan and Shamir was a good exercise in dynamic programming and OpenCV. It is introduced again later in the computational photography PDF below.

Link to seam carving video

Computational Photography in Python

Please scroll through the portfolio below.

Nobu Kim-Toyoda

Software Engineer

フリーランス