Sungyeon Kim

Computer Vision Lab, POSTECH (Pohang University of Science and Technology)

• Advised by Prof. Suha Kwak.
• Recipient of the Alumni Award.
• Dissertation: Towards Retrieval at Scale via Compact Embeddings and Generative Modeling
• Committee: Suha Kwak, Minsu Cho, Seungyong Lee, Jungseul Ok, Bohyung Han
• Research focuses on deep metric learning, multimodal and generative learning, and representation learning.
• Additional research experience includes domain generalization, active learning, and multimodal reasoning with large language and multimodal large language models (LLMs / MLLMs).

Applied Scientist II, Amazon

• Contributed to the research and development of multimodal retrieval and result-verification components for Amazon Lens Live, a real-time video-based visual search system serving millions of users.
• Designed and implemented a client-side query verification module that increased precision and reduced false positives in real-time visual search.
• Built a server-side retrieval validation framework ensuring intent-consistent and trustworthy search results.
• Enhanced the MLLM-based contextual retrieval pipeline to capture reasoning beyond pure visual similarity.

Postdoctoral Researcher, POSTECH

• Conducted postdoctoral research under Prof. Suha Kwak on generative retrieval frameworks for human-centric multimodal video search and compact representation learning.
• Designed a generative retrieval architecture bridging textual and visual semantics for scalable multimodal retrieval.

Applied Scientist Intern, Amazon

• Researched and implemented generative retrieval models for universal multimodal search, exploring decoder-based retrieval generation as an alternative to dense embedding retrieval.
• First-author, GENIUS: A Generative Framework for Universal Multimodal Search, CVPR 2025.
• Collaborated with Xinliang Zhu, Xiaofan Lin, and Muhammet Bastan under Douglas Gray.

Research Collaborator, MIT-IBM Watson AI Lab

• Collaborated with IBM Research on parameter-efficient and unified metric learning for large-scale, cross-domain representation transfer.
• Developed a parameter-efficient transfer learning framework for deep metric learning under distribution shift.
• First-author, Parameter-efficient Transfer for Unified Distance Metric Learning, WACV 2025.

Research Intern, Naver

• Researched with Geonmo Gu and Byungsoo Ko on adapting NLP self-supervised pretraining (e.g., BERT, ELECTRA) to vision for robust representation learning.
• Implemented visual pretraining frameworks inspired by masked language modeling and evaluated them on large-scale image datasets.

International Conference Reviewer

• IEEE Conference on Computer Vision and Pattern Recognition (CVPR) - (Outstanding Reviewer in 2022 and 2025)
• International Conference on Computer Vision (ICCV)
• European Conference on Computer Vision (ECCV)
• International Conference on Machine Learning (ICML)
• International Conference on Learning Representations (ICLR)
• Conference on Neural Information Processing Systems (NeurIPS)
• AAAI Conference on Artificial Intelligence (AAAI)
• IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
• Asian Conference on Computer Vision (ACCV)
• International Conference on Machine Vision Applications (MVA)
• International Conference on Pattern Recognition (ICPR)

International Journal Reviewer

• Transactions on Pattern Analysis and Machine Intelligence (TPAMI), IEEE
• International Journal of Computer Vision (IJCV)
• Transactions on Image Processing (TIP), IEEE