Seokeon Choi

Staff Research Engineer, Qualcomm AI Research · Agentic AI Core System Architecture

I am a Staff Research Engineer at Qualcomm AI Research in San Diego, where I work on hybrid on-device/cloud agentic AI systems, memory-driven personalization, reasoning efficiency, and agentic RAG. My broader research spans generative AI, multimodality, efficient fine-tuning, domain generalization, and human-centric computer vision.

Before moving to the U.S. Agentic AI Core System Architecture team, I worked in Qualcomm Korea's On-device Learning team on diffusion personalization, PEFT/LoRA, low-precision training, and deployment workflows. I received my Ph.D. from KAIST, advised by Prof. Changick Kim, and previously worked with Google Research and Carnegie Mellon University.

CV Google Scholar GitHub LinkedIn DBLP Twitter Email

3.4k+ citations 2k+ GitHub stars 26 international conference papers 7 US patent applications ICCVW 2025 best paper award

News

2026 Two papers accepted to CVPR 2026: memory-efficient DiT fine-tuning and multi-human generation.
2025 Joined the Agentic AI team at Qualcomm AI Research, San Diego.
2025 Received Best Paper Award at the ICCVW LIMIT Workshop.
2025 Selected as a Top Reviewer at NeurIPS 2025.
2025 One paper accepted to NeurIPS 2025 and two papers accepted to ICCV 2025.
2024 One paper accepted to NeurIPS 2024 and one paper accepted to ECCV 2024.

Older news

2023 Promoted to Staff Research Engineer at Qualcomm AI Research.
2023 One paper accepted to CVPR 2023 on single-domain generalization.
2023 One paper accepted to CVPRW 2023 on transfer learning via neural transformation networks.
2022 One paper accepted to ECCV 2022 on test-time adaptation.
2022 Online action detection work accepted to IEEE TPAMI.
2021 Two papers accepted to CVPR 2021 and one paper accepted to ICCV 2021.
2021 Presented at the CVPR 2021 Doctoral Consortium.
2021 Joined Qualcomm AI Research in Seoul.
2021 Research internship with Google Research, TensorFlow Model Optimization.
2020 Visiting researcher at Carnegie Mellon University.
2020 Hi-CMD accepted to CVPR 2020; RLT-DiMP and VOT challenge work appeared at ECCVW 2020.
2019 Domain adaptive object detection work accepted to CVPR 2019; gait recognition work published in TIFS 2019.
2018 Long-term tracking work appeared at ECCVW 2018 and placed in VOT2018-LT challenge results.

Research Topics

My research connects deployment-oriented AI systems with visual and multimodal learning. The arc started from machine perception and human understanding, expanded into generalizable and transferable learning, and now centers on efficient personalized generative models and agentic AI systems for hybrid on-device/cloud environments.

2025 - present

LLM and Agentic AI

Hybrid on-device/cloud agentic systems, memory extraction, user profile lifecycle, preference modeling, and agentic RAG.
Small-language-model reasoning efficiency through test-time adaptation, reinforcement learning, on-device learning, and automated prompt optimization.

2024 - present

Generative AI and Multimodality

Personalization and efficient fine-tuning for diffusion/foundation models [C25][C23][C22][C21][C20][P03]
Multi-human generation, benchmarks, and controllable human synthesis [C26][C24][P02]
Synthetic data and representation generation [C18][C17][C11][C06]

2019 - 2024

Generalizability and Transferability

Domain generalization [C19][C18][C14]
Domain adaptation and test-time adaptation [C16][C06][P01]
Cross-modal learning [C11]
Transfer, contrastive, and few-shot learning [C17][C13]

2015 - 2022

Human Understanding

Person re-identification [C14][C11][P01]
Pose estimation [C12]
Action detection [J02]
Gait recognition / biometrics [J01]

2015 - 2021

Machine Perception

Object detection [C06]
Object tracking [C09][C07][C05]
Object segmentation [C10]
3D reconstruction / multi-view stereo [C15][C03]
Challenge reports [C08][C04]
Low-level vision, image enhancement, and template matching [C02][C01]

Homepage-fit research timeline from 2018 to 2026

Featured Publications

[C26] Ar2Can: An Architect and an Artist Leveraging a Canvas for Multi-Human Generation

Shubhankar Borse, Phuc Pham, Farzad Farhadzadeh, Seokeon Choi, Phong Ha Nguyen, Anh Tuan Tran, Sungrack Yun, Munawar Hayat, and Fatih Porikli

CVPR 2026

paper

[C25] Memory-Efficient Fine-Tuning Diffusion Transformer via Dynamic Patch Sampling and Block Skipping

Sunghyun Park*, Jeongho Kim*, Hyoungwoo Park, Debasmit Das, Sungrack Yun, Munawar Hayat, Jaegul Choo, Fatih Porikli, and Seokeon Choi

CVPR 2026

paper

TBD

[P03] Mix-Opt: Mixed Optimization for Memory-Efficient Personalization of Text-to-Image Diffusion Models

Seokeon Choi*, Sunghyun Park*, Hyoungwoo Park, Jeongho Kim, and Sungrack Yun

Preprint, 2025; extended version of the ICCVW 2025 LIMIT paper

preprint coming soon

[C24] MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans

Shubhankar Borse, Seokeon Choi, Sunghyun Park, Jeongho Kim, Shreya Kadambi, Risheek Garrepalli, Sungrack Yun, Munawar Hayat, and Fatih Porikli

NeurIPS 2025

paper code/data

[P02] From Wardrobe to Canvas: Wardrobe Polyptych LoRA for Part-level Controllable Human Image Generation

Jeongho Kim, Sunghyun Park, Hyoungwoo Park, Sungrack Yun, Jaegul Choo, and Seokeon Choi

Preprint, 2025

paper

[C23] Memory-Efficient Personalization of Text-to-Image Diffusion Models via Selective Optimization Strategies

Seokeon Choi*, Sunghyun Park*, Hyoungwoo Park, Jeongho Kim, and Sungrack Yun

ICCVW 2025 LIMIT Workshop

paperoral presentationbest paper award

[C22] Steering Guidance for Personalized Text-to-Image Diffusion Models

Sunghyun Park*, Seokeon Choi*, Hyoungwoo Park, and Sungrack Yun

ICCV 2025

paper

[C21] ConsNoTrainLoRA: Data-driven Weight Initialization of Low-rank Adapters using Constraints

Debasmit Das*, Hyoungwoo Park*, Munawar Hayat, Seokeon Choi, Sungrack Yun, and Fatih Porikli

ICCV 2025

paper

[C20] Hollowed Net for On-Device Personalization of Text-to-Image Diffusion Models

Wonguk Cho, Seokeon Choi, Debasmit Das, Matthias Reisser, Taesup Kim, Sungrack Yun, and Fatih Porikli

NeurIPS 2024

paper arXiv

[C19] Feature Diversification and Adaptation for Federated Domain Generalization

Seunghan Yang, Seokeon Choi, Hyunsin Park, Sungha Choi, Simyung Chang, and Sungrack Yun

ECCV 2024

paper

[C18] Progressive Random Convolutions for Single Domain Generalization

Seokeon Choi, Debasmit Das, Sungha Choi, Seunghan Yang, Hyunsin Park, and Sungrack Yun

CVPR 2023

paper supp video slide poster

[C17] Neural Transformation Network to Generate Diverse Views for Contrastive Learning

Taekyung Kim, Debasmit Das, Seokeon Choi, Minki Jeong, Seunghan Yang, Sungrack Yun, and Changick Kim

CVPRW 2023

paper supp poster

[C16] Improving Test-Time Adaptation via Shift-agnostic Weight Regularization and Nearest Source Prototypes

Sungha Choi, Seunghan Yang, Seokeon Choi, and Sungrack Yun

ECCV 2022

paper

[J02] Learning to Discriminate Information for Online Action Detection: Analysis and Application

Sumin Lee, Hyunjun Eun, Jinyoung Moon, Seokeon Choi, Yoonhyung Kim, Chanho Jung, and Changick Kim

IEEE TPAMI 2022

paper code

[C15] Just a Few Points are All You Need for Multi-view Stereo: A Novel Semi-supervised Learning Method for Multi-view Stereo

Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, and Changick Kim

ICCV 2021

paper supp

[C14] Meta Batch-Instance Normalization for Generalizable Person Re-Identification

Seokeon Choi, Taekyung Kim, Minki Jeong, Hyoungseob Park, and Changick Kim

CVPR 2021

paper supp code

[C13] Few-shot Open-set Recognition by Transformation Consistency

Minki Jeong, Seokeon Choi, and Changick Kim

CVPR 2021

paper supp code

[C12] MobileHumanPose: Toward Real-time 3D Human Pose Estimation in Mobile Devices

Sangbum Choi, Seokeon Choi, and Changick Kim

CVPRW 2021

paper code

[C11] Hi-CMD: Hierarchical Cross-Modality Disentanglement for Visible-Infrared Person Re-Identification

Seokeon Choi, Sumin Lee, Youngeun Kim, Taekyung Kim, and Changick Kim

CVPR 2020

paper supp code video 1 video 2 poster slide 1 slide 2

[C10] RPM-Net: Robust Pixel-Level Matching Networks for Self-Supervised Video Object Segmentation

Youngeun Kim, Seokeon Choi, Hankyeol Lee, Taekyung Kim, and Changick Kim

WACV 2020

paper supp oral video

[C09] Robust Long-Term Object Tracking via Improved Discriminative Model Prediction

Seokeon Choi, Junhyun Lee, Yunsung Lee, and Alex Hauptmann

ECCVW 2020

paper code video 1 video 2

[C07] Bilinear Siamese Networks with Background Suppression for Visual Object Tracking

Hankyeol Lee, Seokeon Choi, Youngeun Kim, and Changick Kim

BMVC 2019

paperspotlight

[C06] Diversify and Match: A Domain Adaptive Representation Learning Paradigm for Object Detection

Taekyung Kim, Minki Jeong, Seunghyeon Kim, Seokeon Choi, and Changick Kim

CVPR 2019

paper supp code

[J01] Skeleton-based Gait Recognition via Robust Frame-level Matching

Seokeon Choi, Jonghee Kim, Wonjun Kim, and Changick Kim

IEEE TIFS 2019

paper code dataset

[C05] A Memory Model based on the Siamese Network for Long-term Tracking

Hankyeol Lee*, Seokeon Choi*, and Changick Kim

ECCVW 2018

paper code

Publications

International Journals

[J02]Sumin Lee, Hyunjun Eun, Jinyoung Moon, Seokeon Choi, Yoonhyung Kim, Chanho Jung, and Changick Kim, "Learning to Discriminate Information for Online Action Detection: Analysis and Application," IEEE TPAMI, 2022. Link
[J01]Seokeon Choi, Jonghee Kim, Wonjun Kim, and Changick Kim, "Skeleton-based Gait Recognition via Robust Frame-level Matching," IEEE TIFS, 2019. Link

International Conferences

[C26]Shubhankar Borse, Phuc Pham, Farzad Farhadzadeh, Seokeon Choi, Phong Ha Nguyen, Anh Tuan Tran, Sungrack Yun, Munawar Hayat, and Fatih Porikli, "Ar2Can: An Architect and an Artist Leveraging a Canvas for Multi-Human Generation," CVPR 2026. Link
[C25]Sunghyun Park*, Jeongho Kim*, Hyoungwoo Park, Debasmit Das, Sungrack Yun, Munawar Hayat, Jaegul Choo, Fatih Porikli, and Seokeon Choi, "Memory-Efficient Fine-Tuning Diffusion Transformer via Dynamic Patch Sampling and Block Skipping," CVPR 2026. Link
[C24]Shubhankar Borse, Seokeon Choi, Sunghyun Park, Jeongho Kim, Shreya Kadambi, Risheek Garrepalli, Sungrack Yun, Munawar Hayat, and Fatih Porikli, "MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans," NeurIPS 2025. Link
[C23]Seokeon Choi*, Sunghyun Park*, Hyoungwoo Park, Jeongho Kim, and Sungrack Yun, "Memory-Efficient Personalization of Text-to-Image Diffusion Models via Selective Optimization Strategies," ICCVW 2025, oral presentation, best paper award. Link
[C22]Sunghyun Park*, Seokeon Choi*, Sungrack Yun, and Fatih Porikli, "Steering Guidance for Personalized Text-to-Image Diffusion Models," ICCV 2025. Link
[C21]Debasmit Das, Hyoungwoo Park, Munawar Hayat, Seokeon Choi, Sungrack Yun, and Fatih Porikli, "ConsNoTrainLoRA: Data-driven Weight Initialization of Low-rank Adapters using Constraints," ICCV 2025. Link
[C20]Wonguk Cho, Seokeon Choi, Debasmit Das, Matthias Reisser, Taesup Kim, Sungrack Yun, and Fatih Porikli, "Hollowed Net for On-Device Personalization of Text-to-Image Diffusion Models," NeurIPS 2024. Link
[C19]Seunghan Yang, Seokeon Choi, Hyunsin Park, Sungha Choi, Simyung Chang, and Sungrack Yun, "Feature Diversification and Adaptation for Federated Domain Generalization," ECCV 2024. Link
[C18]Seokeon Choi, Debasmit Das, Sungha Choi, Seunghan Yang, Hyunsin Park, and Sungrack Yun, "Progressive Random Convolutions for Single Domain Generalization," CVPR 2023. Link
[C17]Taekyung Kim, Debasmit Das, Seokeon Choi, Minki Jeong, Seunghan Yang, Sungrack Yun, and Changick Kim, "Neural Transformation Network to Generate Diverse Views for Contrastive Learning," CVPRW 2023. Link
[C16]Sungha Choi, Seunghan Yang, Seokeon Choi, and Sungrack Yun, "Improving Test-Time Adaptation via Shift-agnostic Weight Regularization and Nearest Source Prototypes," ECCV 2022. Link
[C15]Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, and Changick Kim, "Just a Few Points are All You Need for Multi-view Stereo: A Novel Semi-supervised Learning Method for Multi-view Stereo," ICCV 2021. Link
[C14]Seokeon Choi, Taekyung Kim, Minki Jeong, Hyoungseob Park, and Changick Kim, "Meta Batch-Instance Normalization for Generalizable Person Re-Identification," CVPR 2021. Link
[C13]Minki Jeong, Seokeon Choi, and Changick Kim, "Few-shot Open-set Recognition by Transformation Consistency," CVPR 2021. Link
[C12]Sangbum Choi, Seokeon Choi, and Changick Kim, "MobileHumanPose: Toward Real-time 3D Human Pose Estimation in Mobile Devices," CVPRW 2021. Link
[C11]Seokeon Choi, Sumin Lee, Youngeun Kim, Taekyung Kim, and Changick Kim, "Hi-CMD: Hierarchical Cross-Modality Disentanglement for Visible-Infrared Person Re-Identification," CVPR 2020. Link
[C10]Youngeun Kim, Seokeon Choi, Hankyeol Lee, Taekyung Kim, and Changick Kim, "RPM-Net: Robust Pixel-Level Matching Networks for Self-Supervised Video Object Segmentation," WACV 2020. Link
[C09]Seokeon Choi, Junhyun Lee, Yunsung Lee, and Alex Hauptmann, "Robust Long-Term Object Tracking via Improved Discriminative Model Prediction," ECCVW 2020. Link
[C08]Matej Kristan, et al. (including Seokeon Choi), "The eighth visual object tracking VOT2020 challenge results," ECCVW 2020. Link
[C07]Hankyeol Lee, Seokeon Choi, Youngeun Kim, and Changick Kim, "Bilinear Siamese Networks with Background Suppression for Visual Object Tracking," BMVC 2019, spotlight. Link
[C06]Taekyung Kim, Minki Jeong, Seunghyeon Kim, Seokeon Choi, and Changick Kim, "Diversify and Match: A Domain Adaptive Representation Learning Paradigm for Object Detection," CVPR 2019. Link
[C05]Hankyeol Lee*, Seokeon Choi*, and Changick Kim, "A Memory Model based on the Siamese Network for Long-term Tracking," ECCVW 2018. Link
[C04]Matej Kristan, et al. (including Seokeon Choi), "The sixth visual object tracking VOT2018 challenge results," ECCVW 2018. Link
[C03]Hankyeol Lee, Seokeon Choi, and Changick Kim, "A Robust Rectification Algorithm for Non-Calibrated Multi-view Images," 3DSA 2018. Link
[C02]Jonghee Kim, Jinsu Kim, Seokeon Choi, Muhammad Abul Hasan, and Changick Kim, "Robust Template Matching Using Scale-Adaptive Deep Convolutional Features," APSIPA ASC 2017. Link
[C01]Seokeon Choi, Jonghee Kim, and Changick Kim, "A Selective Upscaling Method via Super-Resolution Validity Measure," ICCE-Asia 2016. Link

Preprints

[P03]Seokeon Choi*, Sunghyun Park*, Hyoungwoo Park, Jeongho Kim, and Sungrack Yun, "Mix-Opt: Mixed Optimization for Memory-Efficient Personalization of Text-to-Image Diffusion Models," Preprint, 2025.
[P02]Jeongho Kim, Sunghyun Park, Hyoungwoo Park, Sungrack Yun, Jaegul Choo, and Seokeon Choi, "From Wardrobe to Canvas: Wardrobe Polyptych LoRA for Part-level Controllable Human Image Generation," Preprint, 2025. Link
[P01]Youngeun Kim, Seokeon Choi, Taekyung Kim, Sumin Lee, and Changick Kim, "Learning to Align Multi-Camera Domains using Part-Aware Clustering for Unsupervised Video Person Re-Identification," Preprint, 2019. Link

Experience and Education

Qualcomm AI Research, San Diego

Staff Research Engineer, Agentic AI Core System Architecture Team · Nov. 2025 - present

Developing hybrid on-device/cloud agentic AI frameworks for personalized AI systems.
Building pipelines for memory extraction, user preference/profile modeling, profile lifecycle management, and agentic RAG over unstructured data.
Improving small-language-model reasoning efficiency with test-time adaptation, reinforcement learning, on-device learning, and automated prompt optimization.
Designing cost-aware benchmarking and evaluation workflows for automated, scalable agentic AI systems.

Qualcomm AI Research, Seoul

Staff Research Engineer, On-device Learning Team · Nov. 2023 - Nov. 2025

Led research on efficient fine-tuning and on-device learning for LLMs, LVMs, LMMs, and diffusion/foundation models.
Worked on PEFT/LoRA, low-precision training, zeroth-order optimization, checkpointing, diffusion guidance, and deployment-oriented training workflows.
Applied methods to SD1.5, SSD-1B, SANA 1.6B, Llama 3, and Qwen2.5, including consistency-model personalization for faster inference.
Targeted practical generation use cases such as single/multi-human generation, email personalization, and memory-efficient personalization; produced 7 conference papers and 6 patent applications in this period.

Qualcomm AI Research, Seoul

Senior Research Engineer, On-device Learning Team · Sep. 2021 - Nov. 2023

Developed domain generalization and adaptation algorithms for deployment-facing computer vision systems.
Integrated research components into Qualcomm internal toolkits and evaluated robustness under source-free and online adaptation settings.
Worked on source-free object detection for surveillance SDKs; contributed to 4 conference papers and 3 patent applications.

Google Research

Research Intern, TensorFlow Model Optimization Team · Jun. 2021 - Sep. 2021

Studied low-bit quantization for efficient neural network deployment and contributed to TensorFlow Model Optimization Toolkit workflows.
Prepared survey, benchmark, and analysis materials, plus documentation and code submission/review artifacts for research-to-toolkit handoff.

Carnegie Mellon University, Language Technologies Institute

Visiting Researcher, advised by Prof. Alex Hauptmann · Jan. 2020 - Jul. 2020

IITP-supported research visit focused on intelligent surveillance, long-term object tracking, and AI City Challenge work.
Completed six CMU courses and built an AI chatbot project while collaborating with the LTI group.

KAIST

Research Assistant, Computational Intelligence Lab · Mar. 2015 - Jun. 2021

Ph.D. and M.S. research advised by Prof. Changick Kim across person re-identification, domain adaptation/generalization, tracking, pose estimation, gait recognition, and 3D reconstruction.
Served as lab leader and contributed to papers in CVPR, ICCV, WACV, BMVC, IEEE TIFS, and IEEE TPAMI.

Education

Ph.D. in Electrical Engineering, KAIST, 2021. Thesis: Person Re-Identification to Bridge Domain Gaps.
M.S. in Electrical Engineering, KAIST, 2017. Thesis: Robust Model-based Gait Recognition via Candidate Selection and Pose-aware Decision Fusion.
B.S. in Electronic and Electrical Engineering, Sungkyunkwan University, 2015. Summa Cum Laude.

Patents, Collaborations, Awards, and Service

U.S. Patent Applications

[U07] Seokeon Choi, Sunghyun Park, and Sungrack Yun, "Spatiotemporal Attention in Generative Machine Learning Models," U.S. Patent Application No. 19/019,882. Link
[U06] Seokeon Choi, Sunghyun Park, and Sungrack Yun, "Personalized Output Generation in Generative Artificial Intelligence Models," U.S. Patent Application No. 18/959,042. Link
[U05] Wonguk Cho, Matthias Reisser, Debasmit Das, Seokeon Choi, Sungrack Yun, and Fatih Porikli, "Finetuning One or More Neural Networks," U.S. Patent Application No. 18/663,903. Link
[U04] Seokeon Choi, Juntae Lee, and Jaewon Choi, "Privacy-Aware Multi-Modal Generative Autoreply," U.S. Patent Application No. 18/454,456. Link
[U03] Seunghan Yang, Seokeon Choi, Hyunsin Park, Sungha Choi, and Sungrack Yun, "Client-agnostic learning and zero-shot adaptation for federated domain generalization," U.S. Patent Application No. 18/238,998. Link
[U02] Seokeon Choi, Sungha Choi, Seunghan Yang, Hyunsin Park, Debasmit Das, and Sungrack Yun, "Semantic-aware random style aggregation for single domain generalization," U.S. Patent Application No. 18/157,723. Link
[U01] Sungha Choi, Seunghan Yang, Seokeon Choi, and Sungrack Yun, "Test-time adaptation with unlabeled online data," U.S. Patent Application No. 18/086,586. Link

Industry Collaborations

Hanwha R&D Center, Team Leader, Small Object Tracking, Sep. 2020 - Aug. 2021. Small-object/IR tracking, military applications, and DCF-based methods; related ECCVW 2020 long-term tracking paper.
LIG Nex1, Team Member, Image Enhancement and Object Tracking, Mar. 2020 - Aug. 2020. Fast-moving object tracking, IR-based tracking, embedded systems, super-resolution, deblurring, and video stabilization.
Hanwha Systems, Team Leader, AI-Based Object Tracking and Recognition, Mar. 2018 - Feb. 2020. Small-object/IR tracking and IR image generation; related BMVC 2019 spotlight and ECCVW 2018 long-term tracking papers.
ETRI, Team Leader, Image Rectification for Super Multi-View Display, Sep. 2017 - Feb. 2018. Super multi-view display, image rectification, feature extraction/matching, and robust estimation.
SAIT, Team Leader, IR Image Quality Assessment, Mar. 2017 - Aug. 2017. Image quality assessment, face recognition, score-level fusion, and adaptive learning.
Samsung DMC / Samsung Research, Team Member, Super-Resolution and High-Resolution Creation, Mar. 2015 - Feb. 2017. Video super-resolution, metadata-guided texture transfer, texture classification, and perceptual high-resolution creation.

Awards, Honors, and Scholarships

Best Paper Award, ICCVW 2025 LIMIT Workshop, 2025.
Top Reviewer, NeurIPS 2025.
CVPR 2021 Doctoral Consortium Presenter, mentor: Tong Xiao from Meta.
KAIST Chul-Hi Han Augustino Scholarship Foundation academic scholarship, 2021.
Samsung Humantech Paper Award, Silver Prize, top 2%, 2021.
Samsung Humantech Paper Award, Finalist, top 8%, 2020.
Carnegie Mellon University visit funded by Korean government, 2020.
CVPR 2020 AI City Challenge, 16th prize, vehicle re-identification track.
ECCV 2020 VOT2020-LT Challenge, 5th prize.
ECCV 2018 VOT2018-LT Challenge, 3rd prize.
Samsung Advanced Institute of Technology Industry-Academic Scholarship, 2017-2021.
Summa Cum Laude, Sungkyunkwan University, 2015.
Best Work/Thesis Award, Sungkyunkwan University, 2014.
Dean's List, Sungkyunkwan University, 2012 and 2013.
National Science and Technology Scholarship, 2012-2014.
Academic scholarship, Sungkyunkwan University, 2009 and 2012.

Professional Activities

Conference reviewer: NeurIPS (2025, 2026); CVPR (2023, 2024, 2025, 2026); ICCV (2023, 2025); ECCV (2024, 2026); ACM MM (2025); WACV (2025).
Journal reviewer: IJCV; IEEE TNNLS; IEEE TIP; IEEE TIFS; IEEE TMM; IEEE TCSVT; IEEE TGRS; IEEE SPL; IEEE IoT Journal; IET Computer Vision; Neurocomputing.
Teaching assistant: Programming Structure for Electrical Engineering; Image Engineering; Image Understanding; Electronics Design Lab; Signals and Systems.
Invited talks: KAIST, ETRI, and SAIT on domain shift, person re-identification, and long-term object tracking.

Skills

Research: agentic AI, RAG, memory personalization, efficient fine-tuning, PEFT/LoRA, diffusion guidance, domain generalization/adaptation, and human-centric computer vision.
Systems: quantization, low-precision training, checkpointing, model conversion, evaluation, and deployment workflows.
Programming: Python, PyTorch, MATLAB, TensorFlow, and C/C++.
Languages: Korean native; English fluent.