Transforming Human Domain Knowledge into Physical Intelligence.
I am an AI Engineer and System Architect focused on bridging the gap between simulation and real-world deployment. My core interest lies in translating tacit field knowledge, operational constraints, and human context into intelligence that robots and systems can understand.
- Robot Learning & Sim-to-Real: Building simulation intelligence (e.g., Dreamer) that overcomes physical noise in unstructured environments like off-road and agricultural terrain.
- Operational Intelligence: Formalizing human know-how (e.g., safety margins, SOPs) into system logic and reward functions for massive logistics environments.
- Research Infrastructure: Architecting reusable, trustworthy evaluation pipelines for complex, long-context AI tasks.
- Product Systems: Designing AI-native architectures that expose reasoning and own the entire workflow.
- Samsung SDS (2025.02 - 2026.02) | Data Analyst / Logistics System Optimization Intern
- Designed warehouse layout relocation simulations and digitized inbound/outbound processes integrating WMS and SAP data.
- Achieved ~90% reduction in detention costs and 98.7% inventory audit accuracy by aligning system logic with field operational context and SOPs.
- R.O.K. Navy Cyber Operations Center (2021.06 - 2023.01) | Information Security Engineer
- Operated real-time security monitoring and threat detection processes during mission-critical situations.
- Autonomous Off-Road & Agricultural Robotics: Overcoming Sim-to-Real gaps by combining Unity-based RL with LiDAR/YOLO-based navigation for real-world unstructured terrain.
- 2stepED: Architected an automated holistic scoring pipeline (R2C) for evaluating essay creativity, leading to a publication at NAACL 2025.
- InsightFlow: Structured a qualitative research product tying project creation, survey design, and transcript analysis into one seamless AI-native workflow.
- Snappo: Explored a photo-memory platform where users reclaim images through a code-based retrieval flow, focusing on ownership and recall.
- EMNLP (Under Review, 2026): Automated Essay Scoring based on Multi-layer Encoders.
- NAACL Findings (2025): Representation-to-Creativity (R2C): Automated Holistic Scoring Model for Essay Creativity.
- Purdue University Research (2024): Evaluating Privacy Infringement Level in Drone-Captured Images Using Privacy Image Quality Assessment Algorithms (Lead Researcher).
- KIIT Conference (2023): Deep Reinforcement Learning Framework Considering Off-road Driving Environment for Autonomous Farm Robot (Gold Award, Undergraduate Thesis Competition).
"When machines fully understand human context, technology becomes the most powerful leverage to liberate, not replace."