Gangwon Jo

Education

• Mar. 2010 - Feb. 2020

  Ph.D. in Electrical Engineering and Computer Science

  Seoul National University, Seoul, Korea

  Advisor: Prof. Jaejin Lee
  Thesis: High-Level Synthesis of OpenCL Kernels for FPGAs

• Mar. 2006 - Feb. 2010

  B.S. in Computer Science and Engineering

  Seoul National University, Seoul, Korea

  Summa cum laude (GPA: 4.07/4.3)
  Received the SNU College of Engineering Alumni Association's citation for academic excellence

• Mar. 2004 - Feb. 2006

  Daegu Science High School

Work and Research Experiences

• Oct. 2021 - present

  Chief Executive Officer

  Oct. 2020 - Oct. 2021

  Chief Architect

  Moreh, Seoul, Korea

  • We have been developing a PyTorch/TensorFlow-compatible deep learning framework for large-scale clusters and clouds. It generates an intermediate representation at run time and applies a variety of compiler techniques to automatically parallelize and optimize the given workload. It decouples user applications from underlying hardware resources by adding a virtualization layer to enable efficient and flexible resource management. It supports non-NVIDIA accelerators such as AMD GPUs and NPUs.
  • Based on the framework, we have been developing comprehensive AI cloud platform software for various levels of services from IaaS to AIaaS.
  • Based on our capabilities in computing infrastructure, we also develop large AI models (LLMs and multimodal AIs), and computing-aware training and compression techniques.

• Aug. 2016 - Sep. 2020

  Chief Technology Officer

  ManyCoreSoft, Seoul, Korea

  • Led the development of iML, a GPU-accelerated machine learning framework for financial applications such as credit risk management. iML supports novel training algorithms to construct more complex and accurate forest models in a reasonable time. It was actually used to develop a new credit scoring system of Korea Credit Bureau in 2020.
  • Designed and built several high-performance computing systems of companies, research institutes, and universities.

• Mar. 2010 - Jul. 2016

  Research Assistant

  Center for Manycore Programming, Seoul National University

  • Designed and implemented SOFF (SNU OpenCL Framework for FPGAs). SOFF is an open-source framework (= compiler + runtime system + board support package) to run OpenCL applications with data-parallel kernels on FPGAs. It is the first OpenCL framework that correctly compiles all applications in the SPEC ACCEL benchmark suite.
  • Participated in the development of SnuRHAC. SnuRHAC is a CUDA runtime system that provides a single virtual GPU corresponding to multiple GPUs in different nodes. I designed the driver-level cluster unified memory, and developed the memory access pattern analyzer used in SnuRHAC.
  • Maintained SnuCL for six years. SnuCL is an OpenCL framework that naturally extends the original OpenCL semantics to heterogeneous cluster environments. It allows applications to use all the accelerators in multiple nodes as if they were in a single node. Especially, I completely rewrote the source code of the runtime system (version 1.3.X) to support the OpenCL ICD standard and to solve many usability issues.
  • Was the lead architect of the Chundoong supercomputer. Chundoong is the first GPU supercomputer in Korea (ranked #277 in the TOP500 list of November 2012). It used cost-efficient gaming GPUs and adopted a self-made water cooling system. I also developed a highly-optimized LINPACK benchmark implementation for Chundoong.

Honors and Awards

• Nov. 2022

  Innovators Under 35 Korea

  MIT Technology Review

• Aug. 2010 - Feb. 2015

  Graduate Student Scholarship

  Korea Foundation for Advanced Studies

• Feb. 2010

  19th Place

  2010 ACM-ICPC World Finals, Harbin, China

• Mar. 2007

  14th Place

  2007 ACM-ICPC World Finals, Tokyo, Japan

• Mar. 2006 - Feb. 2010

  Presidential Science Scholarship

  Korea Science and Engineering Foundation

• Aug. 2005

  Silver Medal

  17th International Olympiad in Informatics, Nowy Sącz, Poland

Teaching Experiences

• Sep. 2024 - present

  Adjunct Assistant Professor

  Graduate School of Data Science, Seoul National University

• Aug. 2015 - Aug. 2019

  Lecturer

  2019 Accelerator Programming Summer School

  2019 Accelerator Programming Winter School

  2018 Accelerator Programming Summer School

  2018 Accelerator Programming Winter School

  2017 Accelerator Programming Summer School

  2017 Accelerator Programming Winter School

  2016 National Supercomputing Summer School @ SNU

  2016 National Supercomputing Winter School @ SNU

  2015 National Supercomputing Summer School @ SNU

• Mar. 2017 - Jun. 2017

  Teaching Assistant

  SNU 4190.103A Programming Practice (Spring 2017)

• Jun. 2014 - Sep. 2014

  Lecturer

  Multicore Programming Expert Course, Employment-Linked R&D Training Center,
  Small and Medium Business Administration

• Mar. 2010 - Jun. 2014

  Teaching Assistant

  SNU 4190.414A Multicore Computing (Spring 2014)

  SNU 4190.414A Multicore Computing (Spring 2013)

  SNU 4190.409 Compilers (Spring 2012)

  SNU 010.133 Digital Computer Concept and Practice (Spring 2011)

  SNU 010.133 Digital Computer Concept and Practice (Spring 2010)

• Jul. 2006 - Jan. 2009

  Teaching Assistant

  National Training Camp for the Internationl Olympiad in Informatics,
  Korean Institute of Information Scientists and Engineers

Professional Activities

• Nov. 2017 - Dec. 2017

  Artifact Evaluation Committee

  CGO 2018

• Jun. 2017 - Jul. 2017

  Artifact Evaluation Committee

  PACT 2017

• Nov. 2016 - Dec. 2016

  Artifact Evaluation Committee

  PPoPP-CGO 2017

Publications

Conference Papers

1. SnuRHAC: A Runtime for Heterogeneous Accelerator Clusters with CUDA Unified Memory

   Jaehoon Jung, Daeyoung Park, Gangwon Jo, Jungho Park, and Jaejin Lee

   HPDC '21: Proceedings of the 30th International Symposium on High-Performance Parallel and Distributed Computing,

   pp. 107—120, Stockholm, Sweden, June 2021.

   DOI: 10.1145/3431379.3460647

2. SOFF: An OpenCL High-Level Synthesis Framework for FPGAs

   Gangwon Jo, Heehoon Kim, Jeesoo Lee, and Jaejin Lee

   ISCA '20: Proceedings of the 47th Annual International Symposium on Computer Architecture,

   pp. 295—308, Valencia, Spain, June 2020.

   DOI: 10.1109/ISCA45697.2020.00034

3. PIPSEA: A Practical IPsec Gateway on Embedded APUs

   Jungho Park, Wookeun Jung, Gangwon Jo, Ilkoo Lee, and Jaejin Lee

   CCS '16: Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security,

   pp. 1255—1267, Vienna, Austria, October 2016.

   DOI: 10.1145/2976749.2978329

4. A Distributed OpenCL Framework using Redundant Computation and Data Replication

   Junghyun Kim, Gangwon Jo, Jaehoon Jung, Jungwon Kim, and Jaejin Lee

   PLDI '16: Proceedings of the 37th Annual ACM SIGPLAN Conference on Programming Language Design and Implementation,

   pp. 553—569, Santa Barbara, California, USA, June 2016.

   DOI: 10.1145/2908080.2908094

5. Automatic OpenCL Work-Group Size Selection for Multicore CPUs

   Sangmin Seo, Jun Lee, Gangwon Jo, and Jaejin Lee

   PACT '13: Proceedings of the 22nd International Conference on Parallel Architectures and Compilation Techniques,

   pp. 387—397, Edinburgh, Scotland (UK), September 2013.

   DOI: 10.1109/PACT.2013.6618834

6. SnuCL: an OpenCL Framework for Heterogeneous CPU/GPU Clusters

   Jungwon Kim, Sangmin Seo, Jun Lee, Jeongho Nah, Gangwon Jo, and Jaejin Lee

   ICS '12: Proceedings of the 26th ACM International Conference on Supercomputing,

   pp. 341—352, San Servolo Island, Venice, Italy, June 2012.

   DOI: 10.1145/2304576.2304623

7. Performance Characterization of the NAS Parallel Benchmarks in OpenCL

   Sangmin Seo, Gangwon Jo, and Jaejin Lee

   IISWC '11: Proceedings of the 2011 IEEE International Symposium on Workload Characterization,

   pp. 137—148, Austin, Texas, USA, November 2011.

   DOI: 10.1109/IISWC.2011.6114174

Journal Articles

1. Accelerating LINPACK with MPI-OpenCL on Clusters of Multi-GPU Nodes

   Gangwon Jo, Jeongho Nah, Jun Lee, Jungwon Kim, and Jaejin Lee

   IEEE Transactions on Parallel and Distributed Systems (TPDS),

   vol. 26, no. 7, pp. 1814—1825, July 2015.

   DOI: 10.1109/TPDS.2014.2321742

Workshop Papers

1. Memory-Access-Pattern Analysis Techniques for OpenCL Kernels

   Gangwon Jo, Jaehoon Jung, Jiyoung Park, and Jaejin Lee

   LCPC '17: Proceedings of the 30th International Workshop on Languages and Compilers for Parallel Computing,

   pp. 109—126, College Station, Texas, USA, October 2017.

   DOI: 10.1007/978-3-030-35225-7_9

2. OpenCL Framework for ARM Processors with NEON Support

   Gangwon Jo, Won Jong Jeon, Wookeun Jung, Gordon Taft, and Jaejin Lee

   WPMVP '14: Proceedings of the 2014 Workshop on Programming Models for SIMD/Vector Processing,

   pp. 33—40, Orlando, Florida, USA, February 2014.

   DOI: 10.1145/2568058.2568064

3. OpenCL as a Programming Model for GPU Clusters

   Jungwon Kim, Sangmin Seo, Jun Lee, Jeongho Nah, Gangwon Jo, and Jaejin Lee

   LCPC '11: Proceedings of the 24th International Workshop on Languages and Compilers for Parallel Computing,

   pp. 76—90, Fort Collins, Colorado, USA, September 2011.

   DOI: 10.1007/978-3-642-36036-7_6

Posters

1. MAPA: An Automatic Memory Access Pattern Analyzer for GPU Applications

   Gangwon Jo, Jaehoon Jung, Jiyoung Park, and Jaejin Lee

   Poster presentation in PPoPP '17: Proceedings of the 22nd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming,

   pp. 443—444, Austin, Texas, USA, February 2017.

   DOI: 10.1145/3018743.3019034

2. OpenCL as a Unified Programming Model for Heterogeneous CPU/GPU Clusters

   Jungwon Kim, Sangmin Seo, Jun Lee, Jeongho Nah, Gangwon Jo, and Jaejin Lee

   Poster presentation in PPoPP '12: Proceedings of the 17th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming,

   pp. 299—300, New Orleans, Louisiana, USA, February 2012.

   DOI: 10.1145/2145816.2145863

Book Chapters

1. SnuCL: A unified OpenCL framework for heterogeneous clusters

   Jaejin Lee, Gangwon Jo, Wookeun Jung, Hongjune Kim, Junghyun Kim, Yong-Jun Lee, and Jungho Park

   Advances in GPU Research and Practice,

   pp. 23—56, Morgan Kaufmann, September 2016.

   ISBN: 9780128037386

Papers in Korean

1. Techniques to Modify and Execute GPU Code in CUDA Application Binaries

   Jaehoon Jung, Gangwon Jo, and Jaejin Lee

   Korea Computer Congress 2019 (KCC),

   pp. 1101—1103, 2019.

   RISS: 106573136

2. Universal Heterogeneous Programming Environment

   Jungho Park, Gangwon Jo, Wookeun Jung, Hyungmo Kim, Jaehoon Jung, Thanh Tuan Dao, Jiyoung Park, Jungwook Kim, Heehoon Kim, Youngdong Do, PyeongSeok Oh, and Jaejin Lee

   Communications of KIISE,

   vol. 35, no. 10, pp. 18—31, 2017.

   RISS: 103625426

3. Automatic Optimization Methods for Image Processing Programs Using OpenCL

   Jaeho Shin, Gangwon Jo, Ilkoo Lee, and Jaejin Lee

   KIISE Transactions on Computing Practices,

   vol. 23, no. 3, pp. 188—193, 2017.

   RISS: 102918274

4. Hardware and Software Support for Deep Learning

   Wookeun Jung, Jungwook Kim, Thanh Tuan Dao, Jungho Park, Jiyoung Park, Jaeho Shin, Jaehoon Jung, Gangwon Jo, Heehoon Kim, Hyoungwook Nam, and Jaejin Lee

   Communications of KIISE,

   vol. 34, no. 9, pp. 10—20, 2016.

   RISS: 102075344

5. Automatic Optimization Methods for Image Processing Programs Using OpenCL

   Jaeho Shin, Gangwon Jo, Ilkoo Lee, and Jaejin Lee

   Korea Computer Congress 2016 (KCC),

   pp. 1494—1496, 2016.

   RISS: 102087847

6. HPC Technology Trends of Big Data Analyses with Supercomputers

   Gangwon Jo, Junghyun Kim, Thanh Tuan Dao, Wookeun Jung, Jungho Park, Yong-Jun Lee, Jaehoon Jung, Jaeho Shin, and Jaejin Lee

   Communications of KIISE,

   vol. 34, no. 2, pp. 31—42, 2016.

   RISS: 101755688

7. LRC: A Lightweight Communication Library for High Performance Computing

   Jaehoon Jung, Gangwon Jo, Junghyun Kim, Wookeun Jung, and Jaejin Lee

   Korea Computer Congress 2015 (KCC),

   pp. 33—35, 2015.

   RISS: 100640266

8. SnuCL: OpenCL Programming Environment for Heterogeneous Manycore Clusters

   Junghyun Kim, Jungho Park, Gangwon Jo, Thanh Tuan Dao, Jinyoung Joo, Jaehoon Jung, Jungwon Kim, Sangmin Seo, Jun Lee, Jeongho Nah, and Jaejin Lee

   Communications of KIISE,

   vol. 32, no. 5, pp. 66—76, 2014.

   RISS: 100034842

9. Trends on Heterogeneous Supercomputers and a Case Study on the Development of a Supercomputer Chundoong

   Gangwon Jo, Sangmin Seo, Jeongho Nah, Jungwon Kim, Junghyun Kim, Jun Lee, Jungho Park, Yong-Jun Lee, Hongjune Kim, Sooyeon Kang, Jinyoung Joo, Seonmyeong Park, Wookeun Jung, Kihyun Im, and Jaejin Lee

   Communications of KIISE,

   vol. 31, no. 4, pp. 34—41, 2013.

   RISS: 99606073

10. Design and Implementation of Virtual Machines as an Aid in Teaching Computer Concepts

    Jeongho Nah, Gangwon Jo, Sooyeon Kang, Wookeun Jung, and Jaejin Lee

    Korea Computer Congress 2012 (KCC),

    vol. 39, no. 1A, pp. 131—133, 2012.

    RISS: 60197567

11. Current Status and Development Prospects of High Performance Computing Technology

    Jun Lee, Sangmin Seo, Jungwon Kim, Gangwon Jo, Wan Choi, and Jaejin Lee

    The Journal of Korean Institute of Next Generation Computing,

    vol. 8, no. 2, pp. 99—117, 2012.

    RISS: 60098652

12. Implementation of Register Allocator for JavaScript JIT Compiler

    Jeongho Nah, Honggyu Kim, Hongjune Kim, Gangwon Jo, and Jaejin Lee

    2011 KIISE Fall Conference,

    vol. 38, no. 2A, pp. 194—197, 2011.

    RISS: 82732085

13. Measuring JavaScript Performance with a Real World Web Application

    Hongjune Kim, Joo Hwan Lee, Gangwon Jo, and Jaejin Lee

    2011 KIISE Fall Conference,

    vol. 38, no. 2A, pp. 131—134, 2011.

    RISS: 82732048

14. Alias Analysis for JavaScript Program Optimization

    Gangwon Jo, Hongjune Kim, Joo Hwan Lee, Jeongho Nah, and Jaejin Lee

    Korea Computer Congress 2011 (KCC),

    vol. 38, no. 1C, pp. 462—465, 2011.

    RISS: 82666651