Apply on Employer Site

AMD · 19 hours ago

Lead HPC Cluster Network Architect

Austin, TX

Full-time

Hybrid

Lead/Staff

$193K/yr - $289K/yr

AMD is a company focused on building innovative products that enhance computing experiences across various domains. The Lead HPC Cluster Network Architect will design advanced network architectures for AI/ML training and inference systems, collaborating with internal and external partners to optimize GPU communication and ensure seamless operation of AI clusters.

AI InfrastructureArtificial Intelligence (AI)Cloud ComputingComputerEmbedded SystemsGPUHardwareSemiconductor

Growth Opportunities

No H1B

Responsibilities

Designing state of the art cluster network architectures for large AI/ML training and inferencing systems which can be optimized for hyperscale capabilities

Engage with AMD customer base while aligning system and networking architecture

Standardize ethernet network architectures and best practices for GPU-to-GPU communication for deep learning and AI workloads using Infiniband and Ethernet technologies

Co-design new Ethernet technology with AMD partner companies to build the next generation of AI cluster networks

Pioneering system and container networking strategies to facilitate seamless operation and scaling of AI clusters

Developing scalable AI/ML training and inferencing communication network reference architectures for each generation of AMD AI/ML products

Serve as chief network engineer on projects supporting Partner OEM co-design of AI/ML clusters

Participate in design phase of each AMD AI/ML GPU generation by developing cluster communication network architectures and requirements

Collaborate across AMD internal and external partner teams to improve communication performance for AMD AI/ML clusters

Qualification

HPC Cluster Network DesignInfinibandEthernet TechnologiesCluster ManagementNetwork SecurityAutomationWorkload ManagementOSI Network ModelsTCP/IP SuitesAnalytical SkillsProblem-Solving SkillsAttention to DetailVerbal CommunicationWritten CommunicationSelf-Starter

Required

Master's or PhD degree preferred in Mathematics, Statistics, Electrical Engineering, Computer Engineering, or a related computational field; equivalent experience also considered

Preferred

In-depth knowledge and experience with network topologies such as Rail and Fat Tree, and technologies including Infiniband, RDMA, RoCE, NVLINK, and PCIe

Expertise in cluster management, network security, automation, and workload management, along with a solid understanding of OSI network models and TCP/IP suites

Extensive real world experience designing for manageability and operation of hyperscale ethernet networks

Expert in verbal and written communication

Strong analytical/problem-solving skills and pronounced attention to details

Must be a self-starter, and able to independently drive tasks to completion