Description
Hewlett Packard Enterprise is the global edge-to-cloud company advancing the way people live and work. We help companies connect, protect, analyze, and act on their data and applications wherever they live, from edge to cloud, so they can turn insights into outcomes at the speed required to thrive in today’s complex world.
In the HPE Hybrid Cloud, we lead the innovation agenda and technology roadmap for all of HPE. This includes managing the design, development, and product portfolio of our next-generation cloud platform, Green Lake. Working with customers, we help them reimagine their information technology needs to deliver a simple, consumable solution that helps them drive their business results.
We are seeking a Performance Expert to join our Performance Engineering team. The engineer will characterize and optimize the performance of AI solutions across GPUs, servers, storage, networking, system software, and AI frameworks.
Responsibilities:
- Design and execute performance studies for generative AI, machine learning, and other AI workloads.
- Characterize inference and training performance across models, GPUs, precision formats, deployment profiles, and scale configurations.
- Measure latency, throughput, time to first token, inter-token latency, concurrency, scaling efficiency, resource utilization, and cost/performance.
- Evaluate single-GPU, multi-GPU, and multi-node configurations.
- Analyze GPU communication, host-to-GPU bandwidth, PCIe topology, NUMA placement, memory bandwidth, storage, and networking behavior.
- Characterize LLM serving, retrieval-augmented generation, vector database, object detection, speech, multimodal, and agentic AI workloads.
- Develop custom benchmark suites and automation when standard benchmarks do not represent the required workload.
- Identify performance bottlenecks and recommend changes to hardware configuration, BIOS settings, operating systems, drivers, runtimes, frameworks, and applications.
- Establish repeatable test methodologies, baselines, acceptance criteria, and release-performance gates.
- Automate benchmark execution, telemetry collection, result processing, comparison, and reporting.
- Produce clear performance reports, sizing evidence, engineering recommendations, and field-facing guidance.
- Evaluate emerging AI technologies, models, GPU platforms, storage architectures, and distributed inference techniques.
- Collaborate with solution engineering, product management, quality engineering, storage, compute, networking, solution-sizing, field, and technology-partner teams.
Mandatory skills:
- Strong understanding of generative AI and machine learning performance concepts, particularly inference performance.
- Hands-on experience deploying and benchmarking AI or GenAI workloads on GPU-accelerated systems.
- Understanding of key inference metrics like End-to-end latency, Time to first token, Inter-token latency, Token throughput, Request throughput, Concurrency, Batch size, GPU utilization and memory consumption, Scaling efficiency.
- Experience with LLM serving frameworks or inference runtimes such as NVIDIA NIM, vLLM, NVIDIA Triton Inference Server, TensorRT-LLM, Hugging Face, PyTorch, or equivalent technologies.
- Understanding of model precision and quantization formats such as FP32, FP16, BF16, FP8 and NVFP4.
- Working knowledge of tensor parallelism, pipeline parallelism, data parallelism, and multi-GPU or distributed execution.
- Experience with containers and container orchestration, including Docker or Podman and Kubernetes or OpenShift.
- Strong Linux administration, troubleshooting, and performance-analysis skills.
- Proficiency in Python and shell scripting for benchmark automation, workload orchestration, telemetry collection, and result processing.
- Experience using AI performance tools or equivalent benchmark frameworks.
Benefits:
- Comprehensive suite of benefits that supports physical, financial, and emotional wellbeing.
- Investment in career development and growth.
- Unconditionally inclusive work environment.