Lead HPC & AI Infrastructure Architect

Advanced Micro Devices, Inc•San Jose, CA

13h

About The Position

We are seeking a Lead HPC & AI Infrastructure Architect to own the architecture and technical direction of large-scale CPU and GPU platforms across on-premises infrastructure, public cloud, and specialized GPU cloud providers. You will translate scientific computing, engineering, and AI requirements into secure, reliable, and cost-effective platforms for tightly coupled HPC workloads, distributed model training, and inference services. This hands-on leadership role spans design, workload validation, deployment, and lifecycle management. You will guide infrastructure engineering and SRE teams and partner with researchers, AI engineers, cloud providers, and hardware vendors to deliver production-ready platforms.

Requirements

Typically 10+ years in infrastructure, cloud engineering, or distributed systems, including 5+ years designing and delivering production HPC or GPU platforms; equivalent demonstrated experience is welcome.
Ownership of large-scale cluster deployments involving hundreds of nodes or comparable complexity, with measurable improvements in workload performance, reliability, or cost.
Deep expertise in a major cloud platform such as Azure, OCI, or AWS, with experience integrating cloud and on-premises systems. Familiarity with specialized GPU providers such as Crusoe or TensorWave.
Strong Linux troubleshooting skills across CPU/NUMA topology, PCIe, GPU connectivity, drivers, virtualization, containers, and distributed runtime environments.
Hands-on Slurm scheduling and resource governance, plus Kubernetes experience. Understanding of MPI, distributed training, collective communication, and AI inference infrastructure.
AMD Instinct GPUs, ROCm, RCCL, and multi-node PyTorch workloads, including compatibility validation and performance troubleshooting.
HPC networking and storage, including RDMA, RoCE or InfiniBand, topology and oversubscription tradeoffs, parallel I/O, data placement, checkpointing, and enterprise connectivity.
Parallel and shared storage technologies such as Lustre, IBM Storage Scale, BeeGFS, NFS, or VAST, and object storage; network, storage, and collective benchmarks for cluster acceptance.
Terraform or equivalent infrastructure as code, Ansible or comparable configuration management, Python or Bash, and Git-based CI/CD. Familiarity with Azure CycleCloud or AWS ParallelCluster.
Observability and SRE practices using tools such as Prometheus and Grafana, including secure access, resource controls, fault recovery, and production readiness.
Experience establishing observability and SRE practices using tools such as Prometheus and Grafana, and designing security boundaries and resource controls for shared infrastructure.
Ability to lead cross-functional technical work independently, communicate architecture tradeoffs clearly, and guide projects from requirements through production acceptance.
Coordination with hardware and data center teams on rack power, liquid cooling, facility capacity, and other constraints affecting GPU platform design.
Bachelor's degree in Computer Science, Computer Engineering, or a related technical field, or equivalent practical experience.

Nice To Haves

Hands-on experience with AMD Instinct GPUs, ROCm, RCCL, and multi-node PyTorch workloads, including compatibility validation and performance troubleshooting.
Experience with Azure, OCI, Crusoe, or TensorWave GPU environments, and cluster provisioning frameworks such as Azure CycleCloud or AWS ParallelCluster.
Experience with Lustre, IBM Storage Scale, BeeGFS, or VAST, plus network, storage, and collective communication benchmarks used for cluster acceptance.
Experience with LLM serving platforms such as vLLM or SGLang, model parallelism, inference capacity planning, and usage accounting or cost allocation.
Understanding of rack power, liquid cooling, and facility capacity constraints, with experience coordinating platform requirements with hardware and data center teams.

Responsibilities

Own the HPC and AI infrastructure roadmap, reference architectures, and standards for CPU/GPU clusters. Define workload placement across bare metal, virtual machines, Slurm, and Kubernetes.
Lead designs covering CPU/NUMA and GPU topology, accelerator interconnects, networking, storage, and identity. Establish validated hardware, firmware, driver, and runtime baselines.
Architect RDMA-capable fabrics and storage tiers. Validate collective communication, parallel I/O, dataset delivery, and checkpoint performance at target scale.
Design resource sharing through Slurm accounting, fair-share, quotas, QoS, preemption, and tenant isolation. Enable reproducible environments, self-service onboarding, and inference capacity planning.
Standardize provisioning, configuration, and fleet upgrades through infrastructure as code, automation, and CI/CD; manage images, containers, and configuration drift.
Define production acceptance criteria, service-level objectives, and fleet observability. Partner with SRE on fault isolation, job recovery, checkpoint/restart, incident analysis, and capacity planning.
Benchmark HPC and distributed AI workloads; improve scaling efficiency, throughput, inference latency, GPU utilization, and cost efficiency through bottleneck analysis.
Embed federated identity, least-privilege access, tenant and network isolation, secrets management, controlled container privileges, audit logging, and vulnerability management.
Lead architecture reviews, proofs of concept, provider evaluations, and technical escalations. Document tradeoffs, mentor engineers, and align delivery with stakeholders and infrastructure partners.