06 Aug
|
Yotta Data Services Private
|
Panvel
06 Aug
Yotta Data Services Private
Panvel
Job Scope:
As an HPC Admin , you will be responsible for the management and maintenance of GPU Supercomputing clusters on NVIDIA reference architecture. You will ensure optimal performance and uptime of these critical systems, supporting high-performance computing (HPC) requirements.
Job Responsibilities:
Configure, and maintain GPU Supercomputing clusters and associated networking configuration.
Implement and optimize software stacks including MaaS (metal-as-a-service), Job Scheduler (SLURM/PBS), Cloud Orchestration (Kubernetes), and Network Management (NetQ for Ethernet fabric and UFM for InfiniBand).
Conduct performance activities such as debugging, profiling, benchmarking, and tuning of GPU applications on large-scale supercomputing clusters.
Run benchmarking applications from widely used platforms such as MLPerf Training & Inference, AI Training (PyTorch, TensorFlow, NeMo, Megatron-LM), and AI Inference (TensorRT-LLM, Triton Inference Server, vLLM).
Must-Have Skill:
Hands-on experience with NVIDIA GPU, particularly NVIDIA Data Centre GPUs (A100/H100)
Experience in provisioning and managing software stacks like MaaS, Job Scheduler (SLURM/PBS), Cloud Orchestration (Kubernetes), and Network Management (NetQ for Ethernet fabric and UFM for InfiniBand).
Prior experience collaborating with NVIDIA Solution Architect & Engineering teams on large-scale GPU-as-a-service projects.
Familiarity with benchmarking applications from widely used platforms and frameworks, including MLPerf, PyTorch, TensorFlow, NeMo, Megatron-LM, TensorRT-LLM, Triton Inference Server, and vLLM.
Experience in performance engineering, including debugging, profiling, benchmarking, and tuning various GPU applications on large-scale supercomputing clusters.
Valuable to Have Skill:
Knowledge of other HPC technologies and architectures beyond NVIDIA, broadening expertise in the field.
Experience with other cloud platforms and orchestration tools, expanding versatility in deployment environments.
Stro
📌 HPC (Panvel)
🏢 Yotta Data Services Private
📍 Panvel