Hello, I’m Jaeuk Han.
I am an undergraduate in the AI · Software Department at Gachon University and an undergraduate RA at the ISNLP Lab.
My interests include Natural Language Processing (NLP), Information Retrieval (IR), and Reinforcement Learning (RL), with a focus on reproducible experiments and practical system building.

RAG Systems: Korean-focused dual-encoder + reranker pipelines and evidence-grounded Q&A demos.

LLM Agents: Experiments combining LLM reasoning with RL training/evaluation in a Pokémon environment.

Data Engineering: End-to-end pipelines for crawling, parsing, unit/notation normalization, and metadata enrichment.

I’m also interested in how technology can support culture and education. For example, I structured historical manuscripts on Korean traditional liquors (886–1947) into a searchable knowledge base and built interactive demos based on it.
Long-term, I aim to bridge technology and culture as a researcher-developer, contributing to open source and sharing knowledge consistently (GitHub, Kaggle, blog).

• Retrieval-Augmented Generation (RAG)
• Retriever / Reranker
• Reinforcement Learning (RL)

• Assistant Instructor, Taekwang High School - "4th Industrial Revolution & Mathematics" Club
  Delivered 8 sessions (3 hours each) on topics such as linear regression, deep learning, and Arduino. Created slides and hands-on notebooks.

• Undergraduate Research Assistant (RA), ISNLP Lab - Gachon Univ.
  Developed RAG pipelines for competitions (dual encoder, reranker, evaluation automation), conducted MNR/GRPO experiments, and prepared seminar materials.

• taekwang-ai-lectures-2024 - Public archive of high-school AI teaching materials (slides/notebooks).

• Korean Traditional Liquor RAG - Structured late-19th–20th century manuscripts on traditional liquors and built an evidence-grounded Q&A demo. 🍶

• law-demand-dashboard - Matches topic-clustered news/SNS to relevant bills and visualizes results as a static leaderboard.

• 2025 AI Malpyeong Competition (NIKL) - Qualified for the Final Round (Jun 2025 – Sep 2025)
  Built iterative pipelines across retrieval → generation → evaluation, focusing on evidence retrieval quality and strict constraint-following.

• The 2nd Medical AI (MAI) Competition (Korea University College of Medicine) - Solo, Top 6% (Nov 2025 – Dec 2025)
  Designed metric-aligned objectives and ran systematic ablations across backbone/pooling/head variants for representation learning.

For more projects, please check my full repositories list.

Languages

Frameworks & Libraries

Tools