Jeong-min Hong 👋

AI & Robotics undergraduate @ Kwangwoon University
GIST AILAB Summer Research Intern · EASY Lab Undergraduate Researcher

I am an AI & Robotics undergraduate interested in turning perception and intelligence into systems that work in the physical world.

My technical path has expanded from sensor control → Vision AI → Edge AI → Physical AI, and I have worked across assistive robotics, sign-language recognition, edge navigation, embodied AI, balancing-robot hardware and control, and accessibility-aware mobility systems.

Current Interest — Rescue Robotics (구조로봇)
I am currently especially interested in rescue robots that must perceive, navigate, and make robust decisions in uncertain disaster environments.

• GIST AILAB — Summer Research Intern
  Visual Room Rearrangement / Embodied AI research, exploring action-level verification before object-pose-changing actions are executed.

• Gilbut AI — Team Leader
  Accessibility-aware mobility service with personalized route scoring using walking ability, stairs, mobility aids, transfers, obstacles, and weather.

• EASY Lab — Undergraduate Researcher
  Research interests spanning Vision AI, Edge AI, distributed intelligence

At GIST AILAB, I am exploring verification methods for Visual Room Rearrangement: checking whether a planner-selected object-pose-changing action may create a new misplaced object before execution.

At EASY Lab, I reviewed and simulated research ideas around:

• distributed LLM inference in UAV Swarm environments
• model/workload distribution across multiple edge nodes
• Node Failure recovery and fallback structures
• decentralized Edge AI as an alternative to single-server dependence

This research seminar is currently paused while I am participating in the GIST internship.

Following the Sign Angel project and KCI publication, I am continuing exploration of robust Korean sign-language recognition under noise and occlusion, including domain-informed landmark feature engineering.

Real-time safety navigation for pedestrians and mobility-impaired users.

YOLOv11n-seg · NCNN · Raspberry Pi 5 · IPM · OSMnx · Gemini · Edge AI

• YOLO/NCNN-based edge perception
• IPM-based real-world distance estimation
• OSMnx pedestrian routing and risk-aware path recommendation

Emergency sign-language communication system for disaster situations.

MediaPipe · 347D Landmark Features · Bi-LSTM · FastAPI · LLM

• Designed 347-dimensional landmark feature engineering
• Built Bi-LSTM sign-language recognition pipeline
• Accuracy 90.00%, F1-Score 89.11%
• Expanded into a KCI first-author publication

Assistive mobile robot for visually impaired users.

Arduino · IR Sensor · Ultrasonic Sensor · PWM

• Line-tracing based mobility assistance
• Obstacle detection and motor control

Two-wheel balancing robot designed, physically fabricated, and stabilized with IMU-based cascaded PID control.

Autodesk Inventor · DC Gear Motor · IMU · Cascaded PID Control · Mechanical Design

• Designed the complete two-wheel robot hardware structure in Autodesk Inventor
• Created custom base, side, and insert plates around the DC gear-motor drive layout
• Fabricated and assembled the physical robot platform based on the CAD design
• Integrated IMU feedback for real-time attitude estimation
• Implemented the inner loop of a cascaded PID controller and achieved successful two-wheel balance stabilization
• Source repository is private; project details are summarized here without exposing the repository

Terrain-aware exploration robot project.

RealSense RGB-D · IMU · BEV · Costmap · A* · Inventor

• Role: Robotics System Engineer — Vision & Mechanical Design
• RGB-D/IMU terrain perception and mapping
• Costmap/A* planning structure
• Mechanical probe and frame design

AI-powered hairstyle recommendation system.

First Author · Journal of the Institute of Electronics and Information Engineers
Vol. 63, No. 1, pp. 81–88 · Jan. 2026

• MediaPipe-based 347-dimensional feature engineering
• Bi-LSTM sign-language recognition
• Accuracy 90.00% · F1-Score 89.11%
• LLM-based emergency message generation using situational context

• AI 티키타카 Mentor — selected as an AI mentoring program mentor and currently participating in mentoring activities.
• Physics Foundation Mentor — mentoring undergraduate students in foundational physics learning.

Tech stack below follows the technologies listed in my current CV.

• GIST AILAB — Summer Research Internship · Embodied AI / Visual Room Rearrangement
• EASY Lab — Undergraduate Researcher · Vision AI / Edge AI / distributed intelligence
• Gilbut2026 — KT Digital Talent accessibility-aware mobility project
• TERRA128 — terrain-aware Physical AI / robotics project
• HOPE-SL-Partner — Sign Angel project organization