CV
Research Interests
My work focuses on building and optimizing large language models for real-world applications, with an emphasis on efficient and scalable inference, agentic search, and real-time query understanding. I have experience in training small to mid-sized language models for shopping applications and deploying medium-sized models for search systems, with a strong emphasis on latency optimization and system efficiency.
Keywords: Large Language Models (LLMs), LLM Pre-training, Inference Optimization, Agentic Search, Retrieval-Augmented Generation (RAG), Query Understanding
Education
- Ph.D. in School of Computing, Korea Advanced Institute of Science and Technology (KAIST), 2022
- M.S. in School of Computing, Korea Advanced Institute of Science and Technology (KAIST), 2017
- B.S. in Computer Science & Engineering, Sungkyunkwan University, 2016
Work Experience
- Applied Scientist, Amazon (Manhattan, NY), Jul 2023 - Current
- Language Model Pre-training (Stores Foundational AI): Led large-scale pre-training of small to mid-sized language models for shopping applications.
- Agentic Search & LLM Efficiency/Latency Optimization (Prime Video Search): Integrated mid-sized large language models into Prime Video Search for agentic search and real-time query understanding.
- Applied Scientist Intern, Amazon (Bellevue, WA), Mar 2022 - Jun 2022
- Developed an unsupervised sentence representation learning method to more effectively capture semantic similarity between texts.
- Research Scientist Intern, Adobe Research, Sep 2021 - Dec 2021
- Developed word- and relation-level semantic graph representations of documents for domain-specific document retrieval.
Awards
- Naver Ph.D. Fellowship, 2020
Publications
Talks
Semantic Alignment at Various Context Sizes
Talk at Brainlink - State, Limitations, and Future of Large Language Models (LLM), South Korea
Weakly-Supervised Pre-Training for Multi-Hop Retriever
Talk at KISTI, South Korea
Weakly-Supervised Pre-Training for Multi-Hop Retriever
Talk at Naver, South Korea
Weakly-Supervised Pre-Training for Multi-Hop Retriever
Talk at LG AI Research TechTalk, South Korea
Context-Aware Answer Extraction in Question Answering
Talk at Naver, South Korea
Additive Compositionality of Word Vectors
Talk at Naver TechTalk, South Korea
Hierarchical Dirichlet Gaussian Marked Hawkes Process for Narrative Reconstruction in Continuous Time Domain
Talk at NCSoft AI Day, South Korea
Academic Service
- Reviewer: ACL Rolling Review (Oct 2021, Nov 2021, Jan 2022, Apr 2022, June 2022, July 2022, Sep 2022), NeurIPS2022, EMNLP 2022, ICLR 2022, NeurIPS 2021, EMNLP 2021, ACL 2021, ICLR 2021
Scholarship
- Korea National Science&Technology Scholarship (Mar 2012 - Feb 2016)
References
- Prof. Alice Oh, School of Computing, KAIST, alice.oh@kaist.edu