Research & Projects

I care about how technologies land in real institutions. The page below is split into two parts: Research lists the academic projects I have led or contributed to as a researcher (accessibility audit, governance and financial markets, ML interpretability for organizational decisions); Projects lists the engineering and applied ML work I have built — internships, undergraduate projects, and competitions — that supplied the systems-building foundation for the research.

A common sociotechnical thread runs through both: who gets to participate, who is measured, and who is left out.

Research

accessibility audit corporate governance DAO interpretability for high-stakes decisions

Improving Accessibility of an Academic Administration Portal for Students with Disabilities

Junior Ombudsperson (Mar 2025 – Dec 2025)

  • Affiliation: Human Rights Center, Seoul National University (BK21-supported)
  • Achievement: Recognized as an Outstanding Project with Institution-level Impact
  • Links: Korean Report
  • Summary
    • Data: SNU academic-administration portal (mySNU) screens, WCAG conformance observations, and structured interviews with students with disabilities
    • Tools & techniques: WCAG 2.1 conformance audit, stakeholder interviews, institutional reporting through the Human Rights Center
    • Led end-to-end accessibility improvement, including problem scoping, pain-point auditing, and stakeholder communication

AXI for Korean Bond Market Data

Main Researcher (Mar 2024 – Jan 2026)

  • Affiliation: Seoul National University
  • Advisors: Professor Jongsub Lee
  • Achievement: Built an alternative index using both issuance and secondary trading data
  • Summary
    • Data: Korean bond issuance records + secondary-market transaction data (real-time scraped), paired with banking-credit indicators
    • Tools & techniques: Python scraping pipeline, time-series infrastructure, alternative-index construction, market-microstructure analysis
    • Constructed real-time scraping bond transaction and indicators of banking credit for Korean bond market analysis

National Pension Fund Activism: Proxy Voting and Investment Strategies

Main Researcher (Aug 2024 – Jan 2025)

  • Affiliation: Seoul National University
  • Advisors: Professor Sungwook Joh
  • Achievement: Published in Review of Financial Information Studies
  • Links: Paper, Media coverage
  • Summary
    • Data: National Pension Service proxy voting records matched with fund-level investment allocations (panel dataset)
    • Tools & techniques: panel regression, empirical-finance analysis, Stata/R workflow
    • Examined links between NPS proxy voting behavior and investment strategy outcomes

DAO Governance

Research Assistant (Jul 2024 – Sep 2025)

  • Affiliation: Seoul National University and Florida State University
  • Advisors: Professor Jongsub Lee, Jungsuk Han, and Tao Li
  • Links: Paper
  • Summary
    • Data: On-chain DAO governance records — proposals, votes, voter addresses, and treasury flows across multiple DAOs
    • Tools & techniques: on-chain data extraction, participation and concentration measurement, literature synthesis for the ECGI working paper
    • Analyzed decentralized governance participation patterns and decision consequences

Learning Production Process Heterogeneity Across Industries: Implications of Deep Learning for Corporate M&A Decisions

Research Assistant (Jul 2025 – Dec 2025)

  • Affiliation: Seoul National University and Michigan State University
  • Advisors: Professor Jongsub Lee and Hayong Yun
  • Achievement: Built diagnostic visualizations for shared-representation analysis and pruning
  • Links: Paper
  • Summary
    • Data: Trained deep-learning weight matrices from industry-level production-function models used for M&A prediction
    • Tools & techniques: weight-relationship visualization tooling, shared-representation diagnostics, pruning analysis (PyTorch)
    • Developed weight-relationship visualization tools for deep learning model diagnosis

Projects

accessible mobile development computer vision MLOps edge deployment full-stack

Music Coding Education iOS Application for Blind Children

Undergraduate Intern (Dec 2019 – Feb 2020)

  • Affiliation: Human Computer Interaction Lab., Ewha Womans University
  • Advisors: Professor Uran Oh
  • Links: Project page
  • Summary
    • Data: UI prototypes and usability feedback from blind elementary-school students
    • Tools & techniques: iOS VoiceOver APIs, accessible UI patterns, Swift
    • Supported voice-over interaction and implementation of coding education for blind elementary students

Wanted Vehicle License Plate Detection

Intern (Mar 2022 – Jun 2022)

  • Affiliation: Department of Intelligence Automation Service, NEXTLab
  • Achievement: Delivered an industrial prototype to Korea Expressway Corporation
  • Links: Media coverage 1, Media coverage 2
  • Summary
    • Data: Korean vehicle license-plate imagery (deployed to Korea Expressway Corporation)
    • Tools & techniques: LRPNet object detection, TensorFlow Lite on Android, OpenCV C++, NodeJS + Docker Compose FTP pipeline
    • Developed the object detection model based on LRPNet
    • Deployed on android application with TensorflowLite and OpenCV C++
    • Implemented data transfer to an FTP server using event listeners and asynchronous socket communication by NodeJS and Docker Compose

Grape Rating by Three Steps Deep Learning using UAV

MLOps & Fullstack Engineer (Jun 2020 – Feb 2021)

  • Affiliation: Undergraduate Project
  • Achievement:
    • Excellence Prize Issued by Korea Agency of Education, Promotion & Information Service in Food, Agriculture, Forestry & Fisheries (Nov 2020)
    • Presented as the Best Practice at the UIC Barcelona Universitat Internacional de Catalunya
  • Links: Project page, Media coverage 1, Media coverage 2
  • Summary
    • Data: UAV-captured grape-farm imagery across flight trajectories
    • Tools & techniques: three-stage deep-learning pipeline, Ionic PWA + GraphQL, AWS services, PyQt autonomous-drive controller, multi-threaded recording
    • Constructed an autonomous system for grape farms to assess quality and suggest thinning
    • Developed progressive web app (PWA) with Ionic framework and GraphQL
    • Integrated AWS services, UAV path trajectory logic, and multi-threaded recording for model operations
    • Developed a desktop application with PyQT to control autonomous driving and recording the vineyard by multi-threading

AWS CIC Challenge: G-Farm

Computer Vision Engineer (Apr 2021 – May 2022)

  • Affiliation: Agriculture Technology Lab., Sejong University
  • Advisors: Professor Hyunkwon Suh
  • Achievement: Provincial Government Project, Funded by Geumsan County and Applied to Local Farms in Geumsan, Korea
  • Links: Media coverage
  • Summary
    • Data: Greenhouse perilla-leaf imagery collected from Geumsan farms
    • Tools & techniques: Mask R-CNN instance segmentation, leaf-area estimation, AWS-based training workflow
    • Contributed to greenhouse-ready leaf segmentation and leaf-area estimation workflows
    • Calculated perilla leaf area by instance segmentation using MaskRCNN

Counting Strawberry Achene: Deep Learning vs. OpenCV Image Processing

Main Researcher (Jul 2022 – Jun 2023)

  • Advisors: Professor Hyeonkwon Suh
  • Achievement: Presented at the Korea Software Congress
  • Links: Paper
  • Summary
    • Data: Strawberry imagery with achene annotations
    • Tools & techniques: deep-learning object detection vs. OpenCV rule-based processing, side-by-side benchmarking
    • Compared deep-learning detection with OpenCV processing for robust achene counting
    • Benchmarked model-driven and rule-based methods for practical measurement tasks

Flower Detection with Detectron2 and YOLOv5 for Edge Computing in a Strawberry Greenhouse

Research Assistant (Jul 2022 – Jun 2023)

  • Affiliation: Agriculture Technology Lab., Sejong University
  • Advisors: Professor Hyeonkwon Suh
  • Achievement: Presented at the XX CIGR World Congress
  • Links: Paper
  • Summary
    • Data: Strawberry-greenhouse flower imagery under edge-device constraints
    • Tools & techniques: Detectron2, YOLOv5, edge-computing latency/accuracy benchmarking
    • Evaluated Detectron2 vs YOLOv5 under greenhouse edge-computing constraints