Heejae Kim

Email: heejaeee.kim@samsung.com

Education

• Ph.D.  in Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Feb. 2021.

• M.S.  in Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Aug. 2013.

• B.S.  in Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Aug. 2011.

Work Experience

• Staff Researcher, AI Center, Samsung Electronics, Jan. 2025 - Current.

• Staff Researcher, Samsung Advanced Institute of Technology (SAIT), Jun. 2022 - Dec. 2024.

• Postdoctoral Fellow, the School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Feb. 2021 - May 2022.

Research Interest

• Machine learning interatomic potentials (machine learning force fields)

• Large-scale (large-batch) distributed training/inference

• Optimization technique for deep learning

• Explainable AI

• AutoML

Publications

• (preprint) Suhwan Song†, Heejae Kim†, Jaehee Jang, Hyuntae Cho, Gunhee Kim, and Geonu Kim. Scalable Reactive Atomistic Dynamics with GAIA. arXiv preprint arXiv:2509.25798, 2025.

  • code: https://github.com/samsungDS-PoCs/GAIA

  • dataset and model checkpoint: https://huggingface.co/collections/aixsim/gaia

• Yongdeok Kim, Jaehyung Ahn, Myeongwoo Kim, Changin Choi, Heejae Kim, Narankhuu Tuvshinjargal, Seungwon Lee, Yanzi Zhang, Yuan Pei, Xiongzhan Linghu, Jingkun Ma, Lin Chen, Yuehua Dai, Sungjoo Yoo. Breaking MLPerf Training: A Case Study on Optimizing BERT. arXiv preprint arXiv:2402.02447, 2024.

• Geonu Kim, Byunggook Na, Gunhee Kim, Hyuntae Cho, Seungjin Kang, Hee Sun Lee, Saerom Choi, Heejae Kim, Seungwon Lee, and, Yongdeok Kim. Benchmark of machine learning force fields for semiconductor simulations: datasets, metrics, and comparative analysis. in NeurIPS D&B, 2023.

• Heejae Kim, Kyungchae Lee, Changha Lee, SangHyun Hwang, and Chan-Hyun Youn. An Alternating Training Method of Attention-based Adapters for Visual Explanation of Multi-domain Satellite Images. IEEE Access (IF: 3.367), vol. 9, 2021.

• Heejae Kim†, Jiyong Han†, Seong-Hwan Kim, Jisoo Choi, Dongsik Yoon, Minsu Jeon, Eunju Yang, Nhat Pham, Sungpil Woo, Jeongkyu Park, Daeyoung Kim, and Chan-Hyun Youn. IsV2C: An Integrated Road Traffic-Network-Cloud Simulator for V2C Connected Car Services. In IEEE SCC (Core Rank 2017: A), 2017.

• Yuyang Peng, Fawaz Al-Hazemi, Heejae Kim, and Chan-Hyun Youn. Design and optimization for energy-efficient cooperative MIMO transmission in ad hoc networks. IEEE Transactions on Vehicular Technology (IF: 5.978), vol. 66, no. 1, 2017. 

• Yuyang Peng, Fawaz Al-Hazemi, Heejae Kim, and Chan-Hyun Youn. Joint selection for cooperative spectrum sensing in wireless sensor networks. IEEE Sensors Journal (IF: 3.301), vol. 16, no. 22, 2016.

• Heejae Kim, Yun-Gi Ha, and Chan-Hyun Youn. Nash Bargaining Solution-based Datacenter Selection under Cloud Content Delivery Network Environments. IEEE Comsoc MMTC E-Letter, vol. 10, no. 1, 2015. 

• Heejae Kim, Myeongseok Hyeon, Hyungyu Jang, and Chan-Hyun Youn. An Analysis of Virtual Machine Performance for Inter- and Intra- Datacenter Resource Management under Cloud Content Delivery Network Environment. In CAIPT, 2015.

• Myeongseok Hyeon, Heejae Kim, and Chan-Hyun Youn. A Cost-effective VM Offloading Scheme in Hybrid Cloud Environment. In CloudComp, 2015.

• Heejae Kim and Chan-Hyun Youn. Effective Computation Offloading Schemes via Application Partitioning and VM Allocation in Mobile Cloud Environment. In ICFW, 2014.

• Heejae Kim, Yoonki Ha, Yusik Kim, Kyung-no Joo, and Chan-Hyun Youn. A VM Reservation-Based Cloud Service Broker and Its Performance Evaluation. In CloudComp, 2014. 

News

• Oct. 2025. Our MLIP, EquFlash, achieved co-1st on Matbench Discovery. The official leaderboard: https://matbench-discovery.materialsproject.org

• Mar. 2026. Our work with H. Cho et al., “GPU-Accelerated Machine Learning Interatomic Potentials with cuEquivariance and MILAP,” was presented as a poster at NVIDIA GTC 2026.