MTN News interview on Nota’s AI model optimization technology
I was interviewed by MTN News about Nota’s AI model optimization technology, and the segment was broadcast on MTN.
Hancheol Park
I am currently a Senior AI Research Engineer at Nota Inc., where I research and develop model compression and optimization techniques to build efficient AI models. I received my Ph.D. in Computer Science from the School of Computing at KAIST, where my research focused on Natural Language Processing under the supervision of Prof. Jong C. Park.
Email: hancheolp (at) gmail.com
News
Two MoE quantization papers accepted to AdaptFM @ ICML 2026
Two papers on our MoE-specific quantization methods for MoE LLMs were accepted to AdaptFM, an ICML 2026 workshop.
NVIDIA Nemotron Hackathon Seoul
We won 1st place in Track C and were selected as the overall winner at the NVIDIA Nemotron Hackathon Seoul with our data-driven method for MoE model optimization.
Official quantized Solar-Open-100B models released
We released official quantized variants of Solar-Open-100B from the Sovereign AI Foundation Model project using our MoE-specific quantization method.
AWS technical blog on LLM optimization for Inferentia
AWS published a technical blog post introducing our LLM optimization and quantization techniques for AWS Inferentia.
Research Interest
Foundation Models
LLM/VLM pretraining, instruction fine-tuning, retrieval-augmented generation, RAG-aware fine-tuning, vLLM-based serving, custom model architecture integration in vLLM, and model evaluation.
Efficient AI
Efficient LLMs/VLMs, lightweight neural architecture design, quantization, pruning, knowledge distillation, model porting, graph optimization, ONNX/TFLite conversion, and NPU/GPU-aware deployment.
Reliable NLP
Hallucination mitigation, ambiguity detection, uncertainty estimation, selective answering, uncertainty-aware generation, and confidence calibration.
Human-Centric Computer Vision
Image and video understanding, surveillance video analytics, event detection, human activity recognition, and vision systems for user convenience and safety.
Publication
DREAM-MoE: Downstream Routing Error-Aware Margin-Preserving Quantization for Mixture-of-Experts Large Language Models
Hancheol Park, Geonho Lee, Tae-Ho Kim. AdaptFM @ ICML, 2026.
SRA-MoE: Output-Aware Selective Router Alignment for MoE Quantization
Geonho Lee, Hancheol Park, Seunghyun Lee, Jungwook Choi, Tae-Ho Kim. AdaptFM @ ICML, 2026.
Value-and-Structure Alignment for Routing-Consistent Quantization of Mixture-of-Experts Models
Hancheol Park, Geonho Lee, Tairen Piao, Tae-Ho Kim. arXiv preprint, 2026.
Nota AI at GenAI Detection Task 1: Unseen Language-Aware Detection System for Multilingual Machine-Generated Text
Hancheol Park, Jaeyeon Kim, Geonmin Kim, Tae-Ho Kim. GenAIDetect @ COLING, 2025.
Where do LLMs Encode the Knowledge to Assess the Ambiguity?
Hancheol Park, Geonmin Kim. COLING Industry Track, 2025.
Assessing the Answerability of Queries in Retrieval-Augmented Code Generation
Geonmin Kim, Jaeyeon Kim, Hancheol Park, Wooksu Shin, Tae-Ho Kim. arXiv preprint, 2024.
Self-Knowledge Distillation for Learning Ambiguity
Hancheol Park, Soyeong Jeong, Sukmin Cho, Jong C. Park. arXiv preprint, 2024.
Cluster Self-Refinement for Enhanced Online Multi-Camera People Tracking
Jeongho Kim, Wooksu Shin, Hancheol Park, Donghyuk Choi. AI City Challenge Workshop @ CVPR, 2024.
Road Object Detection Robust to Distorted Objects at the Edge Regions of Images
Wooksu Shin, Donghyuk Choi, Hancheol Park, Jeongho Kim. AI City Challenge Workshop @ CVPR, 2024.
Computationally Efficient Decoders for Semantic Segmentation Models
Donghyuk Choi, Hancheol Park, Geonmin Kim, Wooksu Shin. KIISE Transactions on Computing Practices, 2023.
Efficient Semantic Segmentation Models with Weighted Sum-based Feature Fusion Decoders
Donghyuk Choi, Wooksu Shin, Hancheol Park, Ki Min, Youngjun Yoo. KCC, 2023. Best Presentation Award
The First Visual Object Tracking Segmentation VOTS2023 Challenge Results
Matej Kristan et al. (including Hancheol Park). VOTS Workshop @ ICCV, 2023.
Deep Model Compression Also Helps Models Capture Ambiguity
Hancheol Park, Jong C. Park. ACL Long Paper, 2023.
Question-Answering in a Low-resourced Language: Benchmark Dataset and Models for Tigrinya
Fitsum Gaim, Wonsuk Yang, Hancheol Park, Jong C. Park. ACL Long Paper, 2023. Outstanding Paper Award
Addressing the Occlusion Problem in Multi-Camera People Tracking with Human Pose Estimation
Jeongho Kim*, Wooksu Shin*, Hancheol Park*, Jongwon Baek. AI City Challenge Workshop @ CVPR, 2023. * Equal contribution
A Unified Compression Framework for Efficient Speech-Driven Talking-Face Generation
Bo-Kyeong Kim, Jaemin Kang, Daeun Seo, Hancheol Park, Shinkook Choi, Hyoung-Kyu Song, Hyungshin Kim, Sungsu Lim. On-Device Intelligence Workshop @ MLSys, 2023.
Cut Inner Layers: A Structured Pruning Strategy for Efficient U-Net GANs
Bo-Kyeong Kim, Shinkook Choi, Hancheol Park. HAET Workshop @ ICML, 2022.
Constructing a Paraphrase Database for Agglutinative Languages
Hancheol Park, Kyo-Joong Oh, Ho-Jin Choi, Gahgene Gweon. Data & Knowledge Engineering, 2019.
Detection of Non-Standard Meaning Usage with Word Embedding
Huije Lee, Hancheol Park, Wonsuk Yang, Jong C. Park. HCIK, 2018.
Predicting Symptoms of Depression for Social Media Users via Linguistic Patterns
Hoyun Song, Hancheol Park, Wonsuk Yang, Jong C. Park. KSC, 2017.
Neural Theorem Prover with Word Embedding for Efficient Automatic Annotation
Wonsuk Yang, Hancheol Park, Jong C. Park. Journal of KIISE, 2017.
Addressing Low-Resource Problems in Statistical Machine Translation of Manual Signals in Sign Language
Hancheol Park, Jung-Ho Kim, Jong C. Park. Journal of KIISE, 2017.
Enhanced Sign Language Transcription System via Hand Tracking and Pose Estimation
Jung-Ho Kim, Najoung Kim, Hancheol Park, Jong C. Park. Journal of Computing Science and Engineering, 2016.
Neural Theorem Prover with Word Embedding for Efficient Automatic Annotation
Wonsuk Yang, Hancheol Park, Jong C. Park. HCLT, 2016. Best Paper Award
Addressing Low-Resource Problems in Statistical Machine Translation of Sign Language
Hancheol Park, Jung-Ho Kim, Jong C. Park. KCC, 2016. Best Paper Award
Affix Modification-Based Bilingual Pivoting Method for Paraphrase Extraction in Agglutinative Languages
Hancheol Park, Gahgene Gweon, Jeong Heo. BigComp, 2016. AFNLP Best Asian Paper Award
The Correlation between Search Quality and Query Popularity
Jong Myoung Kim, Hancheol Park, Gahgene Gweon, Jeong Hur. BigComp, 2016.
Initiating Moderation in Problematic Smartphone Usage Patterns
Hancheol Park, Gahgene Gweon. CHI Extended Abstracts, 2015.
Measuring Popularity of Machine-Generated Sentences Using Term Count, Document Frequency, and Dependency Language Model
Jong Myoung Kim, Hancheol Park, Young-Seob Jeong, Ho-Jin Choi, Gahgene Gweon, Jeong Hur. PACLIC, 2015.
Sentential Paraphrase Generation for Agglutinative Languages Using SVM with a String Kernel
Hancheol Park, Gahgene Gweon, Ho-Jin Choi, Jeong Heo, Pum-Mo Ryu. PACLIC, 2014.
An Automatic Evaluation Metric for Korean Paraphrase via Semantic Frame
Hancheol Park, Gahgene Gweon, Ho-Jin Choi. KIPS, 2014.
Granted Patent
Method and System for Computationally Efficient High-Performance Object Detection
Hancheol Park. KR Patent No. 10-2931775; Application No. 10-2022-0132409; Granted.
Method and System for Compressing Natural Language Understanding Models via Layer Pruning
Hancheol Park. KR Patent No. 10-2909777; Application No. 10-2022-0132410; Granted.
Technique for Reducing Hallucinated Answers from AI-Based Language Models
Geonmin Kim, Jaeyeon Kim, Wooksu Shin, Hancheol Park. KR Patent No. 10-2916056; Application No. 10-2024-0116795; Granted.
Method and Apparatus for Early Fire Detection
Jeongho Kim, Hancheol Park. KR Patent No. 10-2873458; Application No. 10-2023-0126929; Granted.
Method and Apparatus for Determining Ambiguity in an Input Prompt
Hancheol Park, Geonmin Kim. KR Patent No. 10-2841030; Application No. 10-2024-0127486; Granted.
Method for Estimating Crowd Count, Method for Training a Model for Crowd Count Estimation, and Electronic Device for Performing the Same
Hancheol Park. KR Patent No. 10-2779724; Application No. 10-2023-0145565; Granted.
Method of Lightweighting a Neural Network for Object Recognition, Method of Recognizing an Object Using the Lightweighted Neural Network, and Electronic Device for Performing the Same
Hyungjun Lee, Hancheol Park. US Patent No. 12,443,828; Application No. 18/932,549; Granted.
Method of Lightweighting a Neural Network for Object Recognition, Method of Recognizing an Object Using the Lightweighted Neural Network, and Electronic Device for Performing the Same
Hyungjun Lee, Hancheol Park. KR Patent No. 10-2740182; Application No. 10-2023-0154173; Granted.
Knowledge Distillation Method and System Specialized for Pruning-Based Deep Neural Network Compression
Hancheol Park, Tae-Ho Kim. KR Patent No. 10-2597184; Application No. 10-2022-7041865; Granted.
Statement Reliability Evaluation System and Method Using Commonsense Knowledge and Linguistic Patterns
Jong C. Park, Hancheol Park, Heeje Lee, Wonsuk Yang. KR Patent No. 10-2439165; Application No. 10-2020-0161606; Granted.
System and Method for Constructing Emotion Lexicon by Paraphrasing and Recognizing Emotion Frames
Jong C. Park, Hancheol Park, Jin-Woo Chung, Huije Lee. KR Patent No. 10-2398683; Application No. 10-2017-0106064; Granted.
Method and System for Personality Recognition from Dialogues
Jong C. Park, Hancheol Park, Hoyun Song, Heeje Lee. KR Patent No. 10-2319013; Application No. 10-2020-0011541; Granted.
Apparatus for Detecting Non-standard Meaning Usage of Words, Method for Detecting Non-standard Meaning Usage of Words, and Recording Medium
Jong C. Park, Heeje Lee, Hancheol Park, Wonsuk Yang. KR Patent No. 10-2204341; Application No. 10-2019-0025645; Granted.
System and Method for Communication Training Program over Virtual Reality and Continued Feedback via Mobile Device
Jong C. Park, Jinah Park, Jung-Ho Kim, Youngjin An, Wonsuk Yang, Hancheol Park. KR Patent No. 10-1998482; Application No. 10-2017-0154022; Granted.
Awards & Honors
NVIDIA Nemotron Hackathon Seoul
Track C: Nemotron for Synthetic Data Generation (SDG), 1st Place in Track C and Overall Winner, NVIDIA.
GenAI Content Detection Task 1
Binary Multilingual Machine-Generated Text Detection, Subtask B: Multilingual MGT Detection, 3rd Place, GenAIDetect Workshop, COLING.
NVIDIA AI City Challenge 2024 Track 4
Road Object Detection in Fish-Eye Cameras, 2nd Place out of 53 teams, CVPR.
NVIDIA AI City Challenge 2024 Track 1
Multi-Camera People Tracking, 3rd Place out of 15 teams, CVPR.
NVIDIA AI City Challenge 2023 Track 1
Multi-Camera People Tracking, Top 10, CVPR.
Visual Object Tracking 2023 Challenge
2nd Place out of 47 teams in accuracy, ICCV.
Outstanding Paper Award
Association for Computational Linguistics (ACL).
Best SK AI Partner
SK Tech Summit.
Best Presentation Award
Korea Computer Congress (KCC).
Best Paper Award
Annual Conference on Human & Cognitive Language Technology (HCLT).
Best Paper Award
Korea Computer Congress (KCC).
AFNLP Best Asian Paper Award
IEEE International Conference on Big Data and Smart Computing (BigComp).
Public Highlights
MTN News Interview
Broadcast coverage featuring my interview on Nota’s AI model optimization technology.
Two MoE Quantization Papers Accepted to ICML 2026 Workshop
Media coverage on two MoE quantization papers accepted to an ICML 2026 workshop.
LLM Model Quantization Techniques for AWS Inferentia by Nota AI
Technical blog on LLM quantization techniques for AWS Inferentia.
Two MoE Quantization Papers Accepted to an ICML 2026 Workshop
Nota AI community post about accepted MoE quantization papers.
NotaMoEQuantization: An MoE-Specific Quantization Method for Solar-Open-100B
Technical writing on MoE-specific quantization for Solar-Open-100B.
Solar-Open-100B-NotaMoEQuant-Int4
Public quantized release of Upstage Solar-Open-100B with Int4 weight-only MoE-specific quantization.
Solar-Open-100B-NotaMoEQuant-NVFP4
Public quantized release of Upstage Solar-Open-100B with NVFP4 MoE-specific quantization.
Unseen Language-Aware Detection System for Multilingual Machine-Generated Text
Nota AI technical writing.
KAIST News: ACL 2023 Outstanding Paper Award
KAIST news coverage for low-resource Tigrinya question answering.
NVIDIA GTC 2023 Posters Spotlight: Computer Vision
Public presentation highlight at NVIDIA GTC 2023.