Fabric SOC Architect

Job Title

Fabric SOC Architect

Role Summary

Work on platform and performance architecture for SoC interconnects, cache coherence, and memory subsystems to optimize AI/HPC workloads. The role bridges ML software stacks, compilers, CPU and accelerator microarchitecture, and interconnect fabrics to make data-driven decisions that improve system performance.

Experience Level

Senior (Architect-level preferred). Candidates at various experience levels may be considered; the final level and offer will be determined during the interview process.

Responsibilities

Deliver system-level performance analysis and architecture recommendations that inform SoC interconnect, cache, and memory design.

• Model and analyze on-chip network (NoC) topologies and routing to optimize throughput, latency, and fairness.
• Design and evaluate QoS and traffic-scheduling strategies for heterogeneous workloads.
• Influence cache-coherence protocol and memory hierarchy choices based on workload-driven performance data.
• Develop and validate performance models (pre-silicon and post-silicon) and correlate results with hardware teams.
• Collaborate with compilers, CPU, and accelerator teams to translate workload characteristics into microarchitectural trade-offs.
• Prototype, profile, and iterate on system-level changes using C/C++ tools and benchmarks.

Requirements

Key technical skills and constraints required for the role.

• Deep understanding of NoC topologies, routing algorithms, QoS, and traffic-scheduling techniques.
• Expertise with cache coherence protocols and interfaces such as AMBA CHI and AXI.
• Experience with modern memory and IO technologies: DDR, LPDDR, GDDR, PCIe, CCIX, CXL.
• Proficiency in C/C++ and experience building efficient performance models and analysis tools.
• Strong analytical and data-driven problem-solving skills; ability to turn workloads into architectural requirements.
• Excellent cross-functional collaboration skills; able to work with compiler, CPU, and interconnect teams.
• Nice-to-have: familiarity with ML/AI traffic patterns or experience with formal verification of cache-coherence protocols.
• Employment contingent on eligibility to access U.S. export-controlled technology; may require U.S. citizenship/permanent residency or an export license.

Education Requirements

BS, MS, or PhD in Electrical Engineering (EE), Electrical & Computer Engineering (ECE), Computer Engineering (CE), Computer Science (CS).

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.