GigaIO provides disruptive interconnect technology to extend PCIe outside the server and across racks to achieve game-changing performance, scalability, and composability for Advanced Scale Computing used in AI/ML/DL, advanced analytics, and high-performance computing. The Fabrex™ PCIe Switch interconnect breaks the server boundary to connect dozens to hundreds of heterogeneous compute engines (CPUs, GPUs, FPGAs, ASICs), memory pools, and NVMe storage devices (SSDs) into dynamically composable, high-performance computing systems.
AI-ML Performance Graduate Intern
We are seeking a graduate-level intern to evaluate configuration recipes for deploying Fabrex™ network products in the context of a variety of AI-ML use cases. This person will propose, construct, and evaluate complete customer solutions, characterize the performance of GigaIO’s disruptive Fabrex™ interconnect compared with the performance of legacy solutions, and document reference configurations (recipes) and performance results. This person will report to the VP of Engineering during the period of the internship. He/she will apply his/her expertise with AI-ML use cases, applications, and benchmarks in a variety of market segments. The term of the internship is negotiable and may extend to continued employment, based on mutual desire.
Must Haves:
1) Broad understanding of multiple tiers of systems software; applications, middleware, OS libraries, OS Kernel, OS Drivers.
2) Background in performance analysis and familiarity with a variety of standard AI-ML performance benchmarks (e.g., MLPerf, BERT, ResNet, MLBench, etc.)
3) Familiarity with AI-ML programming using CUDA, multi-GPU NCCL, PyTorch, TensorFlow. 4) Strong scripting skill (Python, Perl, or a Linux shell).
5) C programming skill.
6) Thorough, focused, methodical, with good documentation habits
7) Excellent conversational, written communication, and presentation skills, in English.
Wants:
1) Experience with a network product or system, preferably at the switch level. 2) General knowledge of a variety of interconnect protocols (e.g., PCI-e, InfiniBand, NFS, TCP/IP, and Ethernet)
3) Familiarity with HPC programming including MPI and Libfabric.
Education Requirements:
1) BS and currently working toward MS or PhD in computer science, computer engineering, electronics engineering, mathematics, physics, or similar field.
Send responses to roneill@gigaio.com
6108 Avenida Encinas # B Carlsbad, CA 92011 www.gigaio.com