αAbout
Jaehong Oh.
Robotics engineer and AI researcher working at the boundary between the mathematics of perception and the engineering of embodied systems. Currently Product Owner for RX at ROBOTIS AI — an industrial machine-tending automation platform — and a B.S. in Mechanical Engineering from Soongsil University, Seoul.
Read the full biography — English
What I'm working on
The current main programme is Unified Latent Representation (ULR) : a test of what learned systems share, what makes a representation identical, how learning differs from inference for an observer, and when organization forms. Its current result is deliberately negative — Canon 24 · Ledger 105 records no additional neural-specific ontology in the present registry. SCC is preserved as historical · archived programme, while the original ONN + ORTSF programme is audited · archived programme. ONN's strong higher-order thesis did not survive audit; the scoped boundary, modest surviving signal, and scoped scalar stability result remain public, while system-level certification is still open. These failures and surviving boundaries feed the ULR programme. The original framing remains the published paper of record.
Education
Soongsil University · Seoul
B.S. in Mechanical Engineering, conferred 2026 — focus on Robotics, AI/ML, and Control Systems. Member of the Fluid Mechanics Laboratory and the Intelligent Robotics Laboratory.
Experience
Product Owner, RX · ROBOTIS AI
Full-time. Owns the RX programme end to end — an industrial machine-tending automation platform — from problem definition and scope through architecture decisions, delivery, and the criteria a release has to meet before it ships.
Research Intern · ROBOTIS (Perception Team)
End-to-end autonomous-driving research. Developed and evaluated perception pipelines for autonomous navigation systems.
Research Team Leader · Intelligent Robotics Lab, Soongsil
Led a reinforcement-learning-based Hidden Object Finding project on a robot manipulator — novel algorithms for object discovery in occluded environments, integrating vision, tactile feedback, and predictive reasoning under PyTorch and ROS 2.
Project Team Leader · Fluid Mechanics Lab, Soongsil
Led a Janus-particle synthesis project for bio-pharmaceutical applications. Electrohydrodynamics (EHD) research with 3D-printed Y-shaped microfluidic channels, combined with CFD simulation and experimental validation.
Development Team Leader · Aviation Society Cheonggeumbi
End-to-end development of a 4-axis autonomous flight drone — custom CAD frame (AutoCAD, SolidWorks), PID control, sensor fusion, and a cross-functional mechanical / electrical / software team.
Selected publications
All papers →Published · Int. J. Topol. · 2026 · Partially superseded after audit
Ontology Neural Network and ORTSF: A Framework for Topological Reasoning and Delay-Robust Control
Preprint · arXiv:2508.21272 · 2025
Learning to Assemble the Soma Cube with Legal-Action Masked DQN and Safe ZYZ Regrasp on a Doosan M0609
Preprint · arXiv:2505.03815 · 2025
Towards Cognitive Collaborative Robots: Semantic-Level Integration and Explainable Control for Human-Centric Cooperation
Other selected projects
Industrial Safety Monitoring System · TurtleBot3 + YOLOv5 · 2024
Autonomous patrol robot for real-time detection of helmets, safety vests, and protective eyewear. SLAM-based navigation with dynamic obstacle avoidance (LiDAR + RGB-D); ≥ 95 % detection accuracy; MQTT-based real-time alerting.
Precision Liquid Injection Control System · Fluid Mechanics Lab · 2023
High-precision concentration control for bio-pharmaceutical applications. Load-cell mass measurement with Extended Kalman Filter (0.1 g precision), ROS 2 multi-threaded sensor / control / UI pipeline, modified-Bernoulli feedback for 0.5 % accuracy in target concentration.
RL-based Soma Cube Assembly · Doosan M0609 · 2024
The hardware companion to the published Soma-cube paper — Legal-Action Masking (4,536 → 2,484 actions, 26 % efficiency), ZYZ singularity avoidance (54 % → 96.1 % success), 91 % sim-to-real transfer via Unity domain randomisation.
Technical
- Robotics & control
- ROS 2 · robot manipulation · autonomous systems · SLAM · PID · sensor fusion
- AI & ML
- PyTorch · TensorFlow · reinforcement learning · computer vision · YOLOv5 · deep learning
- Languages
- Python · C++ · MATLAB · Git · Linux
- CAD & mechanical
- AutoCAD · SolidWorks · Autodesk Inventor · 3D printing · CFD
- Tools
- Unity · MQTT · WebSocket · Docker · Intel RealSense · Arduino · Raspberry Pi
Certifications
- CAT (Certified Associate in Technology) Level 1 — Korea Productivity Center, Aug 2024
- Doosan Robotics Bootcamp — Doosan Robotics, 2024
Contact
- jack0682@naver.com
- GitHub
- github.com/jack0682
- arXiv
- arxiv.org / Oh Jaehong
- Location
- Seoul, South Korea