Postdoctoral Researcher, Beyond-G Global Innovation Center
Head of the Beyond-G AI Computation Center, Hanyang University, Seoul, South Korea
I work on AI-native wireless systems for 6G — semantic communication beyond bit fidelity, integrated sensing and communication, and reinforcement learning for O-RAN resource orchestration. I run a DGX H200 cluster and an ARC-OTA testbed, so the work goes from theory to a radio that actually transmits.
Walking a visiting group through the GPU cluster that powers our 6G research.
Six lines of work, each one moving from a theoretical result to something that runs on real infrastructure. The last is still open. The constellation alongside is the problem in one picture: sixteen symbols scattered by channel noise, resolving onto the decision lattice as the receiver acquires lock.
A domain-aware, knowledge-enhanced framework that transmits what a message means rather than reproducing every bit, with physical-layer channel adaptation and data-preserving reconstruction. This was my doctoral thesis and remains the spine of the work.
IEEE TNSE 2026, IEEE OJCOMS 2025, ICT Express 2024Coverage and capacity optimisation inside an O-RAN digital twin, testing static tuning against learned control, and taken through the full safe-automation path: shadow mode, promotion gates and automatic rollback before anything touches a live cell. Built and taught on NVIDIA AODT and VIAVI AIRSG.
IEEE TNSM, submittedA provable Doppler-induced orthogonality-loss gap in semantic-ISAC, validated inside a network digital twin, together with the O-RAN observable the standard currently lacks to detect it. Current work extends this to intent-driven agentic decisions on a UAV-ISAC twin built over ray-traced cities, on a verify-rehearse-commit loop.
IEEE APWCS 2026, Seoul. IEEE TNSE, submittedHierarchical multi-resolution deep Q-networks for physical resource block allocation across B5G heterogeneous networks, combining traffic prediction with interference management. Five patent filings came out of this line.
IEEE TCCN, under reviewOn the UAV-ISAC twin an LLM orchestrator turns an operator's natural-language intent into a typed policy, which deterministic rule verification then checks against the mission contract before anything is rehearsed or committed. The model proposes; it never decides. The same instinct runs through my semantic work, where domain knowledge lets a dual-mode framework carry what text means across a damaged channel instead of reproducing its bits.
IEEE TNSE 2026, vol. 13. IEEE TNSE, submittedNetShield asks whether a 6G network can serve as a certified safety sensor for warehouse robots. A mobile robot's lidar sees only line of sight, so it must crawl past blind spots; base stations that sense people as well as carry traffic could cover what the robot cannot. A safety filter sits between the planner and the wheels and admits a motion command only if the robot could still stop for anyone it has not yet seen, so that if a person does reach it, the robot is already at rest. It runs on a simulated warehouse twin with Nav2 navigation, pedestrians driven by motion-capture data, and a simulated 6G network.
The honest result so far is negative, and more interesting for it: at today's 6G specification the network cannot see well enough near the robot to earn it any extra speed, and the lidar still does the work. That inverted the question into the one the project now answers — what must a network actually provide, in base-station height and count, range resolution, and fusion with the robot's own sensors, before it can certify a robot's surroundings?
In progress, targeting IEEE Transactions on RoboticsRecorded walkthroughs of three systems. Each one is the demo that accompanies the paper named beside it. Nothing downloads until you press play.
Peer-reviewed journal and conference work, with three journal manuscripts currently under review.
Beyond Bit Fidelity: Dual-Mode Domain-Aware Knowledge-Enhanced Semantic Text Communication Framework
Pushing the Boundaries of Urban Wireless Networks: A Semantic-Aware Intelligent Network System for Data-Centric Adaptive Spectrum Management
Navigating the Future of Wireless Networks: A Multidimensional Survey on Semantic Communications
Real-Time AI-Based Informational Decision-Making Support System Utilizing Dynamic Text Sources
Deep Learning Based Multi-Zone Autonomous Valet Parking System Utilizing V2I Communications
Real-Time RL-Based 5G Network Slicing Design and Traffic Model Distribution for V2X and eMBB Services
A Hybrid Power-Rate Management Strategy in Distributed Congestion Control for 5G-NR-V2X Sidelink Communications
Coverage and Capacity Optimization in an O-RAN Digital Twin: Static Tuning or Learned Control? submitted
Intent-Driven Agentic Decisions on a UAV-ISAC Digital Twin over Ray-Traced Cities: Verify, Rehearse, Commit submitted
Advanced DART-PRB: A Hierarchical Multi-Resolution Framework for Efficient Resource Allocation in B5G Heterogeneous Networks under review
Semantic Cyclic Prefix: Transforming Guard Intervals into Intelligent Data Carriers for B5G OFDM Systems under review
Empirical Communication Impairment Thresholds for Safe Cooperative Lane Changing: How Reliable Must V2X Be? accepted
Doppler-Induced Orthogonality Loss in Semantic-ISAC: Provable Gap, Digital-Twin Validation, and the Missing O-RAN Observable
AI-Based Adaptive Beamforming for Wireless Networks: Mitigating NLOS and 3D Beamforming Challenges for Enhanced Coverage
Adaptive AI-Powered CSI Methods for Wireless Networks: Harnessing Advanced Neural Estimation to Overcome NLOS Challenges
5G NR-V2X Communications for Cooperative Resource Management of the Cell Edge User Based on Uu and PC5 Interfaces
Big Data Preprocessing for Real-Time Disaster Situational Awareness by Text Classification
Feasibility Analysis of AI Algorithms for MCS Selection in Wireless Communication Systems
Free apps to download, for Windows, macOS and Linux.
Filed through the Inha University Industry-Academic Cooperation Foundation.
AI-Based Decision Support System Utilizing Dynamic Text Sources
Patent 10-2466559, November 2022Real-Time Reinforcement Learning-Based 5G Network Slicing Design for V2X and eMBB Services
Application 10-2023-0014181, February 2023Reinforcement Learning-Based RAN Slicing Method and System for Dynamic Resource Utilization in 5G V2X, eMBB and mMTC
Application 10-2023-0077967, June 2023Reinforcement Learning-Based Adaptive Slice Resource Management Method and Control Apparatus
Application 10-2024-0135482, October 2024Semantic-Aware Wireless Communication Resource Management Method and Control Apparatus
Application 10-2024-0135483, October 2024Genetic Algorithm Optimization-Based Semantic Intelligent Network System Simulator for Data-Driven Adaptive Spectrum Management
Software C-2024-026063, July 2024Korean government R&D programmes funded through the Ministry of Science and ICT and IITP, coordinated as lab leader at the Mobile Telecommunications Research Lab.
6G core technology for efficient wireless connectivity in three-dimensional space, covering aerial and vertical mobility.
6G wireless access systems targeting ultra-high intelligence and precision network operation.
Next-generation V2X communication and validation technology for intelligent transportation and telematics systems.
Specialised 5G systems for private network deployment, delivered as ICT R&D with international joint research.
Cultivating AI researchers with cross-disciplinary expertise for intelligent industry applications.
From a first software project to running a 6G research computing centre.
Beyond-G Global Innovation Center, Hanyang University
Lead AI-RAN, O-RAN and network digital-twin research. Administer a multi-node NVIDIA DGX H200 and L40S cluster running Slurm and Docker for more than 30 researchers, and maintain the ARC-OTA testbed for ISAC experiments.
Three-day international workshop hosting National Taiwan University of Science and Technology
Designed and delivered the practical track, from CU/DU/RU, RIC and xApps/rApps fundamentals to closed-loop reinforcement-learning control, including a lab that healed a real coverage dead zone from 4 dB back to 10 dB SINR.
Inha University, under Prof. Kyunghi Chang, Chairman of the 6G Forum Korea
Thesis: Beyond Bit Fidelity — a domain-aware, knowledge-enhanced semantic data communication framework. Graduated with a 4.21/4.50 GPA and the Dean's Choice Best PhD Outstanding Research Award.
Inha University
Coordinated five national projects with ETRI, IITP and EUCAST, wrote proposals, delivered technical reporting, and mentored junior researchers who went on to co-author IEEE Access and Sensors papers.
Inha University
Thesis: a real-time AI-based decision support system drawing on dynamic text, speech and image sources.
Competitive awards and research funding.
AgentISAC-6G: a physics-grounded radio vision-language model for ISAC-aware O-RAN orchestration, awarded to Hanyang University for October 2026 through March 2027.
35,000 H100 GPU-hours on NVIDIA Brev, plus one DGX Spark
Inha Talented Award for outstanding research during doctoral study.
Recognised under the Brain Korea 21 Four Program for excellence in artificial intelligence as a promising global AI talent.
Full doctoral funding from Inha University, awarded on academic record and research potential.
Seventy per cent tuition scholarship for master's study.
I am open to postdoctoral and staff scientist positions in 6G, ISAC, O-RAN and AI-native wireless systems, and happy to discuss collaborations or talks.