Siddharth Sharma

GPU Architect at NVIDIA

Milpitas, California, United States

About

Specialties: Graphics architecture profiling and bottleneck analysis. Evaluating and designing algorithms for parallel platforms. Other interests and experiences include - Machine learning techniques, computer vision and video analytics, Gamification and serious games.

Experience

  • GPU Architect at NVIDIA
    Sep 2013 - Present · 12 yrs 10 mos

  • Graduate Research Assistant at North Carolina State University
    Jan 2012 - Aug 2013 · 1 yr 8 mos

    Design and development of parallelization and load balancing techniques for Systems Biology workloads that show irregular parallelism. Typically, design of efficient load sharing technique at intra-GPU, GPU-CPU and inter-GPU level in a distributed GPU cluster system. - Achieved speedups vs CPUs -- 100x+ on One Fermi GPU, 1200x+ of sustained speedup on Fermi-GPU cluster with 16 nodeds. Was Teaching Assistant for Computer systems programming ECE 209 in Spring 12.

  • GPU Architecture intern at NVIDIA
    Jun 2012 - Dec 2012 · 7 mos

    Graphics pipeline - bottleneck analysis CPU/GPU interplay on low power/mobile architectures Developing methodologies to perform effective competitive architecture analysis.

  • Assistant Research Engineer at Seoul National University
    Feb 2010 - May 2011 · 1 yr 4 mos

    Image processing lab@ SNU with Dr. Sang Hwa Lee - Worked on Stereo matching applications and their implementation on GPGPUs - Developed prototype for PTZ camera based intelligent vision surveillance system. See publications for more details. Computer Architecture and Parallel processing lab under Dr. Hyuk Jae Lee - Developed hardware design for Face detection and tracking

  • Engineering Trainee at Wise Automotive, SNU, Seoul
    Sep 2009 - Feb 2010 · 6 mos

    Worked on Vision system development for Automobiles. Project - Fixed point algorithm for camera based blind spot detection in vehicles.