Donghyun Lim

Curriculum Vitae

Donghyun LIM (임동현, 林東炫)

• donghyunismyname@gmail.com
• Research Engineer
• Theoretical Computer Scientist

Skills

C++, CMake, Python, PyTorch, OpenCV, Web, Computer Vision, Theory of Computation

Work Experience

(2024.03~now) Co-founder, Holidayworks

We were too few, I had to do it all

• Software Plugin Development
  • .NET Reverse Engineering
• Data-driven Business Analysis
  • Online Data Scraping
  • Public Data API by the Korean Government
  • Data Analysis with Machine Learning
  • Geolocation Data
• API Integration
  • (Social Networks) KakaoTalk, WhatsApp, Instagram, Twitter
  • (Maps) Kakao Map, Naver Map, Google Map
  • (AI) ChatGPT, ElevenLabs, Naver Clova, StableDiffusion, HuggingFace, KoGPT
• Automated Content Generation
  • Wrote Python scripts for various text/video generation with AI APIs
  • Developed a video generation tool for end users
    • Spreadsheet-based UI
    • React, FastAPI, AI API
• Full Stack Dev
  • Front (HTML, CSS, Javascript, React)
  • Back (Flask, FastAPI, Firestore NoSQL DB)
  • Infra (AWS, GCP, Docker, RunPod, GitHub Pages)

(2021.09~2024.03) Research Engineer, NALBI

Human Motion Capture with Vision AI (2022.04~2024.03)

• Algorithm Implementation in C++
  • Implemented appropriate geometric algorithms for handling rotations in 3D
  • Optimized image processing pipeline by eliminating existing redundancy in C++
  • Devised a novel smoothing algorithm for human motion data sequence
  • Implemented various mathematical operations in C++
  • Designed motion data interpolation pipeline
• On-device ML Inference Engine Deployment
  • Implemented full-tracking inference pipeline involving multiple deep learning models in C++ that runs cross-platform: iOS, Android, and Windows
  • Built Windows DLL for ML deployment with C++ and OpenVINO
  • Built iOS xcframework for ML deployment with C++ and CoreML
  • Built Android AAR library for ML deployment with C++ and Tensorflow Lite
• Miscellaneous
  • Encrypted and decrypted deep learning model weight and model output using AES256
  • Processed human pose datasets for deep learning training
  • Worked in between machine learning researchers and software engineers
  • Helped other researchers and engineers handle mathematical operations

Barcode Decoding from RGB Images (2021.09~2022.03)

• Developed a classical lightweight computer vision algorithm for barcode decoding
• Trained GAN-based image enhancement models using PyTorch
• Implemented deep learning pre/postprocessing algorithms in C++
• Converted PyTorch models into TensorRT models for on-device deployment

Education

(2019~now) Ph.D. candidate in Computer Science, KAIST

• All But Dissertation (not finished)
• Complexity and Real Computation Laboratory

(2017~2019) M.S. in Computer Science, KAIST

• Complexity and Real Computation Laboratory

(2012~2017) B.S. in Computer Science and Mathematics (double major), Chungnam National University

Awards and Honors

• (2018) 5th place, Samsung Collegiate Programming Cup (SCPC) 2018
• (2016) Honorable Mention, ACM-ICPC Asia Daejeon Regional, November 2016
• (2014) Bronze Medal, 33rd University Students Contest of Mathematics, Korean Mathematical Society

Publication

• Donghyun Lim, Martin Ziegler

  Quantitative Coding and Complexity Theory of Continuous Data: Part I – Motivation, Definition, Consequences

  Journal of the ACM

  February 28, 2025

  https://doi.org/10.1145/3705609

• Donghyun Lim, Martin Ziegler

  Quantitative Coding and Complexity Theory of Compact Metric Spaces

  Proceedings of the 16th Conference on Computability in Europe (CiE 2020)

  June, 2020

  https://doi.org/10.1007/978-3-030-51466-2_18

Patents as Inventor, Korean Intellectual Property Office

• (submitted 2023/registered 2024) Registration Number 10-2736905-0000
  • METHOD AND SYSTEM FOR 3D MOTION CAPTURE
• (submitted 2022/registered 2023) Registration Number 10-2519711-0000
  • METHOD AND SYSTEM FOR RECOGNIZING BARCODES BASED ON IMAGES
• (submitted 2017/registered 2018) Registration Number 10-1858178-0000
  • SYSTEM AND METHOD FOR FINDING AN APPROXIMATE SHORTEST PATH CONNECTING TWO PEOPLE IN A SOCIAL NETWORK

Internships

• (2018.07~2018.08) Software Engineering, Elice
  • Customized Jupyter Notebook to capture the interaction between frontend and the language kernel
• (2018.01~2018.02) Software Engineering, Elice
  • Wrote a program that performs syntax analysis on program codes to detect plagiarism
  • Expanded previous work to support not only Python, but other languages such as C, C++, and Java
• (2017.01~2017.02) Chungnam-Purdue National Center for Excellence in Software Program 2017
  • UAV (Unmanned Arial Vehicle, flying drone) detection with acoustic analysis
  • Applied convolutional neural networks to recognize sound fingerprints generated by UAVs

Human Languages

• Korean (native)
• English (professional working proficiency)
• Japanese (JLPT N1, highest level)

Introduce Myself

Donghyun was a graduate student of theoretical computer science, a purely academic field based on mathematical rigor. At one point, he leaped into the industry and started working on practical problems.

Life does not go as planned. Life is full of surprises. Donghyun appreciates serendipity. He is willing to learn anything new. He is open to new adventures and challenges.

He loves optimization on all levels, from high-level algorithms to low-level native code. He has a pathological obsession with making computation faster.

Military Service

Donghyun finished his Korean military service as a research personnel (전문연구요원).

last updated Aug 2024