Thorsten Kurth

Senior Software Engineer at NVIDIA

Zurich, Zurich, Switzerland

About

I am a Gordon Bell Prize Awardee and an energetic and innovative expert for scientific and numerical software. • Strong skills in performance optimization, parallel computing and scalability. • Successfully lead teams and contributed to projects that achieved unparalleled performance of various artificial intelligence/deep learning applications at massive scale, including the first ever deep learning application which reached more than one ExaOp/s peak performance on the Summit supercomputer. Expertise: Targeted performance optimization of numerical applications, distributed deep learning, algorithm development and implementation, programming models for accelerators, statistical analysis of large amounts of data. Experience: Software development, databases, contract management, mentoring. Passion: Using cutting-edge technology to tackle state-of-the-art challenges. Looking for expanding my professional network. mobile: +1 (510) 928-4501 email: [email protected]

Experience

  • NVIDIA (Full-time · 6 yrs 8 mos)
    • Senior Software Engineer
      Jan 2021 - Present · 5 yrs 6 mos

    • Senior Software Engineer
      Nov 2019 - Dec 2020 · 1 yr 2 mos

      In 1999, NVIDIA sparked the growth of the PC gaming market, redefined modern computer graphics, and revolutionized parallel computing. More recently, GPU deep learning ignited modern AI - the next era of computing - with the GPU acting as the brain of computers, robots, and self-driving cars that can perceive and understand the world. • exploring IO performance optimizations for deep learning applications

  • Application Performance Specialist at National Energy Research Scientific Computing Center (NERSC)
    Feb 2016 - Nov 2019 · 3 yrs 10 mos

    The National Energy Research Scientific Computing Center (NERSC) is the primary scientific computing facility for the Office of Science in the U.S. Department of Energy. • lead NESAP for Learning program which aims at improving the productivity of 5 selected artificial intelligence applications by at least 16x on the upcoming Perlmutter HPC System compared to the current Edison HPC System • Technical Representative for the U.S. Department of Energy Path Forward Program: track statement-of-work progress of hardware vendors by reviewing milestone reports as well as organizing and participating in review meetings, participate in milestone acceptance decisions worth $600K-$1.2M each • develop artificial intelligence/deep learning applications for extreme scale and unprecedented performance, including the first ExaOp/s application for which our team was awarded the ACM Gordon Bell Prize 2018 • optimize scientific and data intensive applications for massively parallel architectures, achieving 2.6x average speedup over unoptimized codes on Intel many-core architectures for a broad set of applications, with individual speedups as high as 25x • liaison for U.S. Department of Energy efforts in exploring performance portable programming models, successfully demonstrated that some applications can achieve performance close to non-portable approaches across different contemporary compute architectures • present results at international conferences and write and publish scientific papers as well as performance optimization case studies at docs.nersc.gov/performance/case-studies

  • Postdoc at Lawrence Berkeley National Laboratory
    Nov 2013 - Jan 2016 · 2 yrs 3 mos

    World renowned research lab of the U.S. Department of Energy. • lead developer in the research collaboration: develop and improve lattice QCD HPC applications • develop domain specific language for automatically generating fast lattice QCD tensor contraction code, achieving about 30x speedup over comparable implementations • analysis of multiple terabytes of structured data • algorithm development and performance optimizations for HPC computations • performance optimizations with CUDA in HPC calculations • I/O performance optimization for massively parallel applications using HDF5 • write and publish scientific papers and present results at international conferences

  • University of Wuppertal (Wuppertal, Germany)
    • Research Associate
      Jul 2007 - Oct 2013 · 6 yrs 4 mos

      The University of Wuppertal is a dynamic university with an interdisciplinary teaching and research profile and strong international collaborations. • precise determination of physical observables through Monte Carlo calculations, many regarded benchmark calculations by the community • HPC application and algorithm development and optimization, leading to 10x speedup over previous code • maintain computer systems and cluster computers • write and publish scientific papers and present results at international conferences

    • Student Assistant
      Nov 2005 - Jun 2007 · 1 yr 8 mos

      • maintain cluster computer ALiCENext • tutor courses with 15-25 students

  • Summer Intern at CERN
    Apr 2005 - Jun 2005 · 3 mos

    The European Organization for Nuclear Research (CERN), is one of the world's largest and most respected centers for scientific research. • Software QA: testing and bug reporting of grid middleware "gLite"