Interconnect and Compute Architect

Job Title

Interconnect and Compute Architect

Role Summary

Design and define next-generation CPU networking and interconnect architectures for datacenter and emerging robotics/automotive AI/ML devices. The role focuses initially on datacenter networking while seeding medium- to low-power robotics/automotive designs.

Work across hardware, software, and systems teams to translate networking requirements into architectural tradeoffs and implementation guidance.

Experience Level

Mid-level. Candidates at various seniority levels will be considered; the appropriate level and offer will be determined during the interview process.

Responsibilities

Contribute to architecture, specification, and cross-team direction for CPU networking and interconnects.

• Define and architect die-to-die and on-chip/off-chip interconnects for datacenter and robotics/automotive targets.
• Model and analyze network performance, including latency, buffering, and throughput tradeoffs.
• Map Ethernet and RDMA networking concepts onto on-chip and off-chip fabrics.
• Drive architectural decisions across forwarding, buffering, modeling, and RTL implementation constraints.
• Collaborate with hardware, firmware, software, and systems teams to refine networking requirements and specifications.
• Help specify medium- to low-power robotics/automotive designs after initial datacenter work.

Requirements

We separate core (must-have) qualifications from desirable (nice-to-have) experience below.

Must-have:

• Practical knowledge of Ethernet network architecture and performance modeling.
• Experience with die-to-die interfaces and related protocols and tradeoffs.
• Familiarity mapping Ethernet networking concepts to on-chip and off-chip fabrics.
• Experience with datacenter scale-up interconnects such as UALink, NVLink, or Broadcom SUE.
• Experience with scale-out RDMA protocols (RoCE, InfiniBand, or similar).
• Ability to reason about forwarding, buffering, performance modeling, and RTL constraints to guide architecture.
• Proven cross-disciplinary collaboration with hardware, software, and systems teams.
• Employment is contingent on eligibility to access U.S. export-controlled technology; offers may depend on citizenship, permanent residency, or ability to obtain export licenses.

Nice-to-have:

• Experience with AI/ML-specific networking protocols and implementations.
• Experience building low-latency networks for HPC and Ethernet topologies.
• Experience targeting robotics or automotive low-power compute/sensor ingest designs.

Education Requirements

Not specified.

About the Company

Company: Tenstorrent

Headquarters: Austin, Texas, United States

Tenstorrent is a technology company focused on designing innovative computing solutions. They are known for their expertise in the development of advanced hardware, including ASICs and SoCs, aimed at enhancing performance and efficiency in various applications.