Lead AI Interconnect Architect for Scale-Out Data Centers

Sandisk; Milpitas (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Sandisk seeks an AI Interconnect Architect to define high-speed networking for AI inference infrastructure, including servers, racks, and chips. The role emphasizes bandwidth, power efficiency, scalability, and optimized transport protocols with PCIe, CXL, UALink, and Ultra Ethernet focus.

You will design scale-out architectures, collaborate with SoC, package design, and software teams, and shape industry standards through performance modeling.

Qualifications

• Master's or PhD in EE/CE/CS as required.
• 10–15 years developing interconnect technologies including transport, link protocols, switching fabrics, QoS.
• Familiarity with Fat tree, Leaf-Spine, Torus, and Meshed topologies.
• Deep knowledge of PCIe, CXL, NVLink, UALink, Ethernet and related protocols.
• Experience with GPU/accelerator clusters and data center infrastructure.

Responsibilities

• System Architecture & Design: chip-to-chip interconnects and switched fabrics for AI/ML scale-out.
• Performance Optimization: analyze bandwidth, latency, power, area, and reliability trade-offs.
• Technology development and standards: contribute to industry specifications.
• Cross-functional Collaboration: work with SoC, package design, and software teams.

Skills

AI Hardware Architecture

Data center infrastructure

GPU clusters

Interconnect knowledge

Performance modeling

Education

Master's or Ph.D. in Electrical Engineering, Computer Engineering, or Computer Science

Tools

PCIe

CXL

NVLink

UALink

Ethernet

Ultra-Ethernet

Serial links

Job description

Sandisk seeks an AI Interconnect Architect to define high-speed networking for AI inference infrastructure, including servers, racks, and chips. The role emphasizes bandwidth, power efficiency, scalability, and optimized transport protocols with PCIe, CXL, UALink, and Ultra Ethernet focus.

You will design scale-out architectures, collaborate with SoC, package design, and software teams, and shape industry standards through performance modeling.