Research Experience

Siebel School of Computing and Data Science. University of Illinois Urbana-Champaign
Undergraduate Research Scholar, Accessibility and AI Project | Sep 2024 – May 2025
Mentor: Professor Lawrence Angrave

🔗Github Repository | 📋View URS Poster: Optical Flow | 📋View ISUR Poster: Multi-Method Scene Detection

• Led a research project to improve accessibility by converting lecture videos into book formats (PDFs), benefiting diverse learners and enhancing educational outcomes.
• Implemented a Python-based pipeline to segment videos into chapters and extract representative images, producing well-structured book content.
• Developed AI-based solutions using open-source AI models (e.g., Ollama Vision, VGG-16) to address challenges such as slide transitions, animations, and annotations.
• Transitioned beyond traditional methods like pixel-based analysis and SVM by integrating advanced AI tools to improve segmentation accuracy and efficiency.
• Designed accessible outputs optimized for usability and readability, adhering to universal design principles.
• Explored CNN models to detect significant scene changes and incorporating OCR for textual extraction from slides.
• Evaluated the quality and usability of generated book outputs through user feedback and iterative design improvements.
• Presented the project at the Undergraduate Researh Sympoisium (URS) at UIUC poster session under the title "Enhancing Keyframe Extraction in Lecture Videos Using Optical Flow."

Siebel School of Computing and Data Science. University of Illinois Urbana-Champaign
The Impact of LLM-Based Tools on User Engagement and the Quality of Q\&A Discussions on Tech Forums | Jan 2025 – May 2025
Mentor: Professor Eshwar Chandrasekharan

🔗Github Repository | 📋View Research Poster | 📝View Project Report

• Conducted a computational social science research project to investigate how the rise of Large Language Model (LLM)-generated content has influenced user behavior and perceived answer quality on Stack Overflow from 2022 to 2024.
• Designed and implemented a scalable data pipeline using shell scripts and Python to extract, chunk, and preprocess 100,000+ Stack Overflow comments, ensuring balanced and representative samples from each year.
• Engineered time-offset features for comments by grouping data by thread and calculating their delay from the first response, enabling a fine-grained analysis of comment timing and engagement.
• Integrated GPT-detection scores through fuzzy matching techniques (using RapidFuzz), aligning LLM-generated comment predictions with high-scoring responses to study overlap between helpfulness and AI content.
• Visualized trends between comment timing, score, and predicted LLM-use to determine whether users value early, human-written answers more than delayed or LLM-generated responses.
• Presented research poster at Undergraduate Research Poster Session at UIUC.

University of Suwon, South Korea
Research Assistant | Apr 2019 – Mar 2020
Mentor: Dr. Woo Rin Lee

• Conducted research targeting the CD44/Beta-catenin pathway as a potential therapeutic strategy for gastric cancer patients.
• Performed and analyzed experiments, including agarose electrophoresis, PCR, and cell maintenance, to evaluate the efficacy of the proposed strategy.
• Published findings in the Journal of Emerging Investigators (JEI), Harvard University. Read the publication.