Academic CV

Yonghyun Gwon mathewgwon@gmail. com | Portfolio | GitHub | LinkedIn EDUCATION Kyung Hee University — Seoul & Yongin , South Korea Expected Feb 202 7 B. Eng. in Software Convergence | GPA: 4.3 3 / 4.5 | Full Tuition Scholarship — 8 semesters PUBLICATIONS Conference Papers • KunHo Heo, SuYeon Kim, Yonghyun Gwon , Youngbin Kim, MyeongAh Cho. "ParTY : Part - Guidance for Expressive Text - to - Motion Synthesis." IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026. • Yonghyun Gwon , KunHo Heo, YoungKwan Cho, SuYeon Kim, HaYoung Lee, MyeongAh Cho. "Word - Motion Cross Attention for Fine - Grained Text Guided Human - Human Interaction Motion Generation." Korea Software Congress (KSC), 2025. Posters & Research Demonstrations • Jaehyun Byun, Byunghoon Kang, Yonghyun Gwon , Hongsong Choi, Yunseo Do, Eunho Kim, Sangkeun Park, Seungjae Oh. " CrossGaussian: Enhancing Remote Collaboration through 3D Gaussian Splatting and Real - T ime 360° Streaming." ACM UIST 2025 Adjunct ; IEEE ISMAR 2025 Adjunct . RESEARCH EXPERIENCE Visual Science Lab (VSLab), Kyung Hee University Jul 2025 - Feb 2026 Undergraduate Research Assistant | Advisor: Prof. MyeongAh Cho • Proposed Word - Motion Cross Attention (WMCA) for text - driven human - human interaction motion generation, improving InterHuman text - motion retrieval by +2.3 / +3.4 / +3.2 points on R - Precision Top - 1 /2/3 while maintaining MM - Dist. • Extended the WMCA study into a completed undergraduate thesis evaluating sequence - level, time - indexed, and event/state interaction representations through conditioning, candidate - selection, and residual - actionability analyses. The results motivated represe ntations that explicitly encode interaction timing, involved body parts, participant roles, and event order. • Contributed to ParTY’s qualitative evaluation through matched - prompt and matched - seed comparisons across HumanML3D , MoMask, ParCO, and ParTY, assessing part - level prompt adherence and failure cases; transformed the initial renders into publication - ready, part - aware figures through substantially improved framing, body - part distinction, and rendering quality. Interactive Technology & Methodology Lab (HCI Lab), Kyung Hee University Sep 2024 - Dec 2024 Research Contributor | Advisor: Prof. Seungjae Oh • Led the initial Cross - Reality and Reality - Virtuality Continuum literature survey and translated early research exploration into experimental scenarios for 3DGS - based remote collaboration, contributing to CrossGaussian. • Prototyped the 360° capture - to - 3DGS and Unity collaborative - viewing workflow, testing both networked view synchronization and direct TCP/UDP data sharing across two machines. SkinHaptics Project | HCI Lab, Kyung Hee University Apr 2024 - Jul 2024 Research Contributor | PI: Prof. Seungjae Oh • Built the Unity VR experiment environment used in the CHI 2025 study, implementing Euler - based vertex deformation, distance - based attenuation, and Oculus hand - skeleton gesture triggers for skin - softness perception experiments. WORK EXPERIENCE FunctionBay / MetaMotion X — Sim ulation / Digital Twin Developer Jul 2024 - Aug 2025 Pangyo Techno Valley, South Korea • Developed and integrated replay and event - logging features within a large existing Unity codebase for a defense - linked simulation, resolving file I/O failures, incomplete writes, and multi - PC log - collection issues. • Designed and implemented network - synchronized multi - client simulation systems, maintaining consistent runtime behavior by debugging state transitions and client desynchronization during integration. • Authored technical specifications and architecture documentation for production features, translating implementation details into clear requirements and supporting unit and integration testing, bug reproduction, and fix verification. M 5ers - Backend Intern Nov 2024 - Jan 2025 Remote & San Jose, CA, USA • Built a Node.js/Express backend and C UART control process for an FPGA - based mining - hardware prototype, connecting register read/write APIs to low - level command routing, response formatting, and integration testing. Republic of Korea Army - UNIX Server Operator May 2021 - Nov 2022 The 1st Logistic Support Command, South Korea • Operated UNIX servers for a military C4I intranet, supporting security checks, log audits, backups, access control, network troubleshooting, and Log4j incident - response procedures under command guidance.

PROJECTS GenMotion2Physics May 202 6 - Present Tools : Python, MuJoCo , PyTorch • Built a MuJoCo retargeting pipeline that maps reference and generated human motion to an articulated SMPLSim humanoid using coordinate/root correction and damped Jacobian IK, achieving approximately 1.7 cm mean joint - position error. • Developed free - root StablePD /PPO motion tracking under gravity and contact, then tested specialist policies, distillation, and rehearsal for acquiring new motions while retaining existing tracking capability. • Compared ground - truth and generated motions using the same humanoid and controller. Ground truth passed 3/3 versus 8/30 generated motions; some failures were already visible in the source or retargeted motion, while others emerged only during physics track ing. Open - Source Contributions • Microsoft / ONNX Runtime ( Merged , PR#29879 ): Fixed a CUDA tensor - reduction bug that interpreted values by byte size rather than actual numeric type, causing silently incorrect arithmetic results, and added broad regression coverage. • N VIDIA/ CUTLASS (Merged , PR#3394 ): Fixed an FP8 matrix - multiplication indexing error that could select incorrect scaling values when GPU thread blocks were reordered, producing incorrect outputs or invalid memory access under specific layouts . • N VIDIA/ CCCL (Merged , PR#9861 ): Fixed a compatibility defect that prevented a CUB (CUDA primitives library) parallel algorithm from working with CUDA container iterators and added cross - compiler regression coverage . AI - Based Multiplayer Latency Predictor Nov 2024 - Dec 2024 Tools: C++, Python, Unreal Engine 5, PyTorch, ONNX/NNE • Led a UE5 latency - compensation project by reframing nonlinear vehicle prediction as motion - delta regression and training a PyTorch Residual+GLU network on custom vehicle - motion features . • Integrated the ONNX model through Unreal Engine 5 NNE for asynchronous 10 Hz inference inside a custom movement component, improving drift - trajectory tracking over constant - velocity prediction. TEACHING & MENTORING Course Mentor — AI and Game Programming Mar 202 6 – Jun 202 6 • Designed and developed the base Unity ML - Agents framework used by all students in the course as the starting point for thei r reinforcement - learning term projects, and mentored teams during implementation. Game Development Club – Study Leader & Mentor 2023, 2024 • Led Unreal Engine C++ and Unity study sessions, mentoring student developers in multiplayer programming, gameplay systems, and engine workflows. HONORS & ACTIVITIES 1 st Place — UNIJAM Game Jam (Smilegate & UNIDEV) 2024 2 nd Place — Immersive Tourism Content Contest (Korea Tourism Organization) 2024 2 nd Place — University Portfolio Competition (Kyung Hee University) 2024 Additional University Awards — 2 nd place, 3 rd place, and Creativity Award in Software and Artistic - Software Competitions 2024 Exhibitor — G - STAR Game Exhibition (Korea Association of Game Industry) 2023, 2024 Selected Participant — NexonJam Competition (28 programmers selected nationwide ) 2024 TECHNICAL SKILLS Advanced: C/C++, C#, Unity, Unreal Engine Proficient: Python, PyTorch, ONNX/NNE, Node.js/Express Familiar: OpenCV, JavaScript, HTML/CSS Tools: Git, Perforce (P4V), Jira, Azure DevOps, Redmine Languages: English ( TOEFL iBT 5.0/6.0 ) , Korean (native), Japanese (conversational)