I am currently working as a Research Associate at the Artificial Intelligence and Mathematical Learning Lab (AIML@K), Department of Mathematics, Korea University.
Believing that the ultimate paradigm of artificial intelligence will manifest as a fully converged system of robotics and AI—akin to Jarvis—I am profoundly interested in its foundational architectures, specifically Agentic AI and Large Action Models (LAM).
Outside the lab, I am discreetly conducting practical experiments on building automated AI-driven quantitative trading systems to generate passive income streams.
Interests
- Algebraic Representation Learning
- Hallucination Mitigation & Detection
- Trustworthy & Reliable AI
Education
M.S. in Mathematics(Mathematical Data Science), 2026
Korea University
B.S. in Mathematics, 2024
Korea University
Research Focus
My ultimate goal is to research and develop next-generation AI models that are verifiable and reliable, fully grounded in mathematical rigor.
Research Goal
My ultimate goal is to develop highly reliable and verifiable AI models built upon solid mathematical foundations.
Research Experience
Core AI R&D: Algebraic Number Theory for LLM Reasoning
- Phase 1 (2025.07 - 2025.12) | Numbers Already Carry Their Own Embeddings
Proposed a novel embedding mechanism leveraging the Adele Ring and $p$-adic algebraic structures to resolve the preservation loss of numerical semantics in conventional LLM tokenizers. This work was accepted at the NeurIPS 2025 Workshop MATHAI. - Phase 2 (2026.01 - 2026.05) | Prime Fourier Embeddings: A Principled Basis for Modular Arithmetic
Introduced the concept of Pontryagin Dual Space to construct a mapping pipeline that projects numbers onto $(\cos, \sin)$ on the complex plane, experimentally demonstrating its intrinsic efficiency in learning the Chinese Remainder Theorem (CRT). This work was accepted at the ICML 2026 Workshop AI4MATH. - Phase 3 (2026.05 - PRESENT)
Currently developing a highly practical, production-ready embedding module that can be seamlessly integrated into arbitrary LLMs as a plug-in component.
Internship at SK Magic
Served as a Data Scientist Intern within the AI Business Development Team at SK Magic. Spearheaded the R&D and design of a hybrid optimization algorithm for the optimal spatial placement of auxiliary air quality sensors paired with the autonomous driving air purifier (NAMUHX).
KCC 2024 and Hallucination
Classified hallucination phenomena in Question-Answering (QA) generation tasks into five fine-grained taxonomies. Evaluated the capabilities and limitations of standard natural language metrics (BLEU, METEOR, ROUGE) in detecting specific types of hallucinations. This work was published at KCC 2024.
AI Grand Challenge 2023
Built a robust synthetic dataset generation and refinement pipeline for multi-answer question answering tasks utilizing the OpenAI API (GPT-3.5), directly contributing to the lab consortium’s final 7th place achievement.