About
I am a Senior Research Engineer in the Multi-modal Development Team at NHN Corporation, where I lead research and development on generative AI for music and visual content creation. Prior to NHN, I was a Research Engineer at Hyundai Motor Company, working on camera–LiDAR fusion for autonomous driving perception. I received my B.S. and M.S. from Sungkyunkwan University.
My research interests lie in understanding the fundamental principles underlying relationships across visual and other modalities. I am particularly interested in representation-level alignment beyond explicit cross-modal matching, and in using such aligned representations to build controllable generative models toward multimodal world foundation models. Related essays can be found on my blog.
Publications
-
PhysTrace: Interpretable Temporal Evaluation of Physical Consistency in Video Generative ModelsPreprint, 2026Independent -
PADS-TAL: Padding Annealed Diffusion Sampling in Text-Aware Latent Space for Robust and Diverse Text-to-Music GenerationInternational Conference on Machine Learning (ICML), 2026PresenterIndependent -
-
-
Analysis on Accelerating Object Detection in Edge Device with Half-Precision Floating PointKorea Computer Congress (KCC), 2018PresenterAward -
Optimizing Trim Operations by In-Storage ProcessingKorea Software Congress (KSC), 2017PresenterAward
Patents
- W. K. Jung, T. K. Yoo, K. H. Kim, "Method and System for Retrieval-Augmented Prompt Enhancement for Music Generation"KR App. 10-2026-0151891
- T. K. Yoo, W. K. Jung, K. H. Kim, "Method and System for Generating Audio Source Incorporating User-Specified Melody Using a Diffusion Model"KR App. 10-2025-0134391US App. 19/810,102
- K. H. Kim, T. K. Yoo, "Method and System for Generating Music Using Diffusion Model Based on BPM Measurement"KR App. 10-2025-0044627
- T. K. Yoo, K. H. Kim, "Method and System for Generating Music Using Diffusion Model Based on Genre-Style Structured Information"KR App. 10-2025-0035765
- J. H. Lim, H. C. Choi, J. H. Kim, T. K. Yoo, "Object Recognition Apparatus and Vehicle Having the Same"US 2024/0046658
- J. H. Kim, H. C. Choi, J. H. Lim, T. K. Yoo, "Apparatus and Method for Object Recognition"US 2024/0104935
- J. T. Kim, T. K. Yoo, D. Y. Kim, "Neural Network Accelerator Using Double-Stage Weight Sharing"KR 10-2519210
Experience
-
NHN Corporation 2023 – PresentSenior Research Engineer, Multi-modal Development Team, AI Tech LabGenerative AI R&D for music and visual content — diffusion-based text-to-music, text-to-image for webtoon character creation. End-to-end production service development with 4 patent applications.
-
Hyundai Motor Company 2020 – 2023Research Engineer, Autonomous Driving SW TeamCamera–LiDAR fusion for 3D object detection and road information recognition. Proposed the company's first fusion-based perception method; deployed in production vehicles. 2 registered patents.
-
Sungkyunkwan University 2018 – 2020M.S., Electrical and Computer EngineeringDeep learning optimization for embedded models (pruning, quantization). Proposed Double-Stage Weight Sharing, reducing model parameters by 80% while maintaining ~90% accuracy.
Skills
- Expertise Multimodal Learning, Generative Models, Deep Learning Optimization, Computer Vision
- Modality Video, Vision, Text, Audio, Point Cloud, ...
- Programming Python, C/C++, CUDA, Golang, JavaScript, Verilog, ...
- Deep Learning PyTorch, PyTorch Lightning, TensorFlow, Diffusers, DeepSpeed, AWS SageMaker, ...
- Infrastructure Linux, Docker, Kubernetes, FastAPI, gRPC, ...
- DL Optimization TensorRT, CUDA, OpenCL, ACL, Nsight Profiler, ...