Hyunkyung Bae

AI Scientist · LLM Training & Inference
LinkedIn Google Scholar

AI scientist working on large-scale LLM training and efficient inference. Currently at NYU's Courant Institute researching memory-efficient multimodal inference; previously 3+ years at LG AI Research building enterprise AI agents on in-house foundation models.

Experience

LG AI Research
NLP Researcher (LLM) · Seoul
Mar 2022 — Nov 2024
ChatEXAONE — Enterprise AI Agent Website · Report
  • Adapted in-house foundation models to document-heavy enterprise workflows via post-training and domain-specific data curation.
  • Trained 3B–30B+ models across multi-node GPU clusters with DeepSpeed ZeRO-2/3, SLURM, and data/model parallelism.
  • Extended long-context capability from 4K to 32K tokens for long and multi-document workflows.
  • Designed evaluation-driven data refinement loops, turning failure cases into better synthetic and instruction-tuning data.
Scientific Paper Conversational QA Agent
  • Built the query-rewriting component, resolving ambiguous queries through context-aware reformulation.
  • Generated ~100K scientific-paper QA instances via a data-augmentation pipeline, contributing to Dialogizer (EMNLP 2023) and IterCQR (NAACL 2024).

Publications

Reducing Peak Memory Usage for Modern Multimodal Large Language Model Pipelines↗
H. Bae*, J. Kim*  ·  ACL Findings 2026
LLMs Can Be Easily Confused by Instructional Distractions↗
Y. Hwang, Y. Kim, J. Koo, T. Kang, H. Bae, K. Jung  ·  ACL 2025
SWITCH: Studying with Teacher for Knowledge Distillation of Large Language Models↗
J. Koo, Y. Hwang, Y. Kim, T. Kang, H. Bae, K. Jung  ·  NAACL Findings 2025
IterCQR: Iterative Conversational Query Reformulation with Retrieval Guidance↗
Y. Jang, K. Lee, H. Bae, H. Lee, K. Jung  ·  NAACL 2024
Dialogizer: Context-aware Conversational QA Dataset Generation from Textual Sources↗
Y. Hwang*, Y. Kim*, H. Bae, H. Lee, J. Bang, K. Jung  ·  EMNLP 2023
Injecting Comparison Skills in Task-Oriented Dialogue Systems for Database Search Results Disambiguation↗
Y. Kim*, Y. Hwang*, J. Shin, H. Bae, K. Jung  ·  ACL Findings 2023
Relevance Similarity Scorer and Entity-Guided Reranking for Knowledge-Grounded Dialog Systems
H. Bae, M. Lee, A. Kim, H. Lee, C. Lee, C. Park, D. Kim, K. Jung  ·  AAAI Workshop 2021

Education

New York University — M.S. Computer Science2025 — 2027
Courant Institute · GPA 3.61/4.0
Seoul National University — M.S. ECE2020 — 2022
Advisor: Kyomin Jung · GPA 3.55/4.3
POSTECH — B.S. Chemical Engineering2010 — 2015
GPA 3.80/4.3 · magna cum laude

Skills

Training — PyTorch, HF Transformers, DeepSpeed ZeRO-2/3, TRL, SLURM, distributed training, instruction tuning, preference optimization
Inference & Agents — vLLM, TensorRT, LangChain, LangGraph, RAG, long-context modeling, document QA
Data & Eval — synthetic data generation, domain data curation, query rewriting, model evaluation, failure-case analysis
Programming — Python, C++