Youngeun Kwon

I am a Senior Deep Learning Performance Engineer at NVIDIA, mainly contributing to NVIDIA's software ecosystems for LLM training (NeMo/NeMo-RL/Megatron-Bridge/Megatron-Core/TransformerEngine/etc). My role is to make deep learning workloads run faster on NVIDIA GPUs, focusing on GPU/system architecture and entire software stacks for deep learning. In the past, I earned my Ph.D. in electrical engineering at KAIST under the guidance of Professor Minsoo Rhu. My primary research interests were computer systems/architecture for deep learning and emerging applications.

• NVIDIA, Santa Clara, CA, USA [Jun. 2024 ~ Present]

• Deep Learning Architect @ E2E training performance team

  • Maximizing datacenter-scale pre-training and post-training performance for LLMs

• Manager: Nitin

• KAIST (Korea Advanced Institute of Science and Technology), Daejeon, South Korea [Sep. 2018 ~ Feb. 2024]

• Ph.D. in Electrical Engineering

• Thesis: Hardware and Software Systems for Accelerating Large-Scale Deep Learning Recommendation Models

• Vertically Integrated Architecture Research Group (VIA)

• Adviser: Prof. Minsoo Rhu

• POSTECH (Pohang Science and Technology), Pohang, South Korea [Mar. 2013 ~ Aug. 2017]

• • B.S. in Computer Science and Engineering

• Graduation Research: Unsupervised Object Detection

• Jiwon Lee, Yunjae Lee, Youngeun Kwon, and Minsoo Rhu, "Characterization and Analysis of the 3D Gaussian Splatting Rendering Pipeline," IEEE Computer Architecture Letters (CAL), 2024

  • [Paper]

• Juntaek Lim, Youngeun Kwon, Ranggi Hwang, Kiwan Maeng, Edward Suh, and Minsoo Rhu, "LazyDP: Co-Designing Algorithm-Software for Scalable Training of Differentially Private Recommendation Models," The 29th ACM International Conference on Architectural Support for  Programming Languages and Operating Systems (ASPLOS-29), San Diego, CA, April 2024

  • Acceptance Rate: 20% (193 among 921)

  • [Paper]

• Youngeun Kwon and Minsoo Rhu, "Training Personalized Recommendation Systems from (GPU) Scratch: Look Forward not Backwards," The 49th IEEE/ACM International Symposium on Computer Architecture (ISCA-49), New York, NY, June 2022

  • Acceptance Rate: 16% (67 among 400)

  • [Paper][Slide][Talk]

• Yunjae Lee, Youngeun Kwon, and Minsoo Rhu, "Understanding the Implication of Non-Volatile Memory for Large-Scale Graph Neural Network Training", IEEE Computer Architecture Letters (CAL), Jul. 2021

  • [Paper]

• Youngeun Kwon, Yunjae Lee, and Minsoo Rhu, "Tensor Casting: Co-Designing Algorithm-Architecture for Personalized Recommendation Training," The 27th IEEE International Symposium on High-Performance Computer Architecture (HPCA-27), Seoul, South Korea, Feb. 2021 

  • Acceptance Rate: 24% (63 among 258)

  • [Paper][Talk]

• Ranggi Hwang, Taehun Kim, Youngeun Kwon, and Minsoo Rhu, "Centaur: A Chiplet-based, Hybrid Sparse-Dense Accelerator for Personalized Recommendations," The 47th International Symposium on Computer Architecture (ISCA-47), Valencia, Spain, June 2020

  • Acceptance Rate: 18% (77 among 421)

  • [Paper]

• Bongjoon Hyun, Youngeun Kwon, Yujeong Choi, John Kim, and Minsoo Rhu, "NeuMMU: Architectural Support for Efficient Address Translations in NPUs," The 25th ACM International Conference on Computer Architectural Support for Programming Languages and Operating Systems (ASPLOS-25), Lausanne, Switzerland, Mar. 2020

  • Selected for IEEE Micro Top Picks Honorable Mention ("IEEE Micro - Top Picks From the 2020 Computer Architecture")

  • Acceptance Rate: 18% (86 among 476)

  • [Paper]

• Youngeun Kwon, Yunjae Lee, and Minsoo Rhu, "TensorDIMM: A Practical Near-Memory Processing Architecture for Embeddings and Tensor Operations in Deep Learning," The 52nd IEEE/ACM International Symposium on Microarchitecture (MICRO-52), Columbus, OH, Oct. 2019

  • Selected for IEEE Micro Top Picks Honorable Mention ("IEEE Micro - The 2019 Top Picks in Computer Architecture")

  • Acceptance Rate: 22% (79 among 344)

  • [Paper]

• Youngeun Kwon and Minsoo Rhu, "A Disaggregated Memory System for Deep Learning," IEEE Micro, Special Issue on Machine Learning Acceleration, Sep/Oct., 2019

  • [Paper]

• Youngeun Kwon and Minsoo Rhu, "Beyond the Memory Wall: A Case for Memory-centric HPC System for Deep Learning", The 51st IEEE/ACM International Symposium on Microarchitecture (MICRO-51), Fukuoka, Japan, Oct. 2018

  • Acceptance Rate: 21% (74 among 351)

  • [Paper][Lightning talk]

• Youngeun Kwon and Minsoo Rhu, "A Case for Memory-Centric HPC System Architecture for Training Deep Neural Networks", in IEEE Computer Architecture Letters (CAL), vol. 17, no. 2, pp. 134-138, July-Dec. 2018. 

  • [Paper]

• Minsoo Rhu, Mike O'Connor, Niladrish Chatterjee, Jeff Pool, Youngeun Kwon, and Stephen W. Keckler, "Compressing DMA Engine: Leveraging Activation Sparsity for Training Deep Neural Networks," The 24th IEEE International Symposium on High-Performance Computer Architecture (HPCA-24), Vienna, Austria, Feb. 2018 

  • Acceptance Rate: 20% (54 among 260)

  • [Paper] [Talk]

• Honorable Mention in IEEE Micro Top Picks 2020 [Mar. 2021]

• The 27th Samsung Humantech Paper Award [Feb. 2021]

  • Gold Prize (1st place in the Computer Science and Engineering track)

• Honorable Mention in IEEE Micro Top Picks 2019 [Mar. 2020]

• KAIST Breakthroughs, Fall 2020 Vol. 14 [Fall 2020]

• Global Ph.D. Fellowship, National Research Foundation of Korea (NRF) [Mar. 2019 ~ Feb. 2024]

• • Research Topic: Memory-centric Architecture for Deep Learning Acceleration

  • Acceptance Rate: 15.2% (216 among 1,423)

• Best presentation award (1st place) in the graduation research project at POSTECH CSE department [July 2017]

• • Research Topic: Unsupervised Object Detection

• Scholar of the National Academic Excellence Scholarship for Science and Engineering, Korea Student Aid Foundation (KOSAF) [Mar. 2013 ~ Dec. 2016]

• Neural network acceleration system and operating method thereof

• Inventor: Minsoo Rhu, Youngeun Kwon, and Yunjae Lee

• Applications: KR (2019), US (2020), CN (2020)

• Registered: KR (2022)

• Current Assignee: KAIST

• High performance computing system for deep learning

• Inventor: Minsoo Rhu, and Youngeun Kwon

• Applications: KR (2018), US (2019)

• Registered: KR (2022), US (2025)

• Current Assignee: Samsung Electronics

Extended Review Committee

• High-Performance Computer Architecture (HPCA): 2024

• International Symposium on Microarchitecture (MICRO): 2025

Journal Reviewer

• IEEE JETCAS: 2024

• Computer Architecture Letters (CAL): 2023

• Applied Artificial Intelligence: 2023

• ETRI Journal: 2020

Discussion Chair

• PeRSonAI tutorial at ISCA: 2020

• KAIST, Daejeon, South Korea

  • Teaching Assistant

    • EE595: Special Topics in Electrical and Computer Engineering <Parallel Computer Architecture> [2021-Spring]

    • EE312: Introduction to Computer Architecture [2021-Fall, 2020-Fall, 2020-Spring]

    • EE209: Programming Structure for Electrical Engineering [2022-Spring, 2019-Fall]

• POSTECH, Pohang, South Korea

  • Teaching Assistant

    • CSED311: Computer Architecture [2018-Spring]