# AI/ML Silicon Accelerator Memory Architect

Broadcom; San Jose (CA); On-site; USD 200,000 - 320,000; Full time

Benefits offered by this job

Discretionary annual bonus

Equity grants

401(k) matching

Job summary

Broadcom in San Jose invites you to lead memory subsystem and data transport architecture for AI/ML accelerators. Own end-to-end path from memory tech through controller, on-die network, DMA engines, and scratchpads, across multiple silicon generations with power-performance trade-offs.

You will model, simulate, and influence choices with internal teams and memory/IP vendors, delivering scalable solutions that meet AI workloads demands for latency and bandwidth.

Qualifications

• 18+ years in ASIC/SoC architecture focused on memory systems and data movement for HPC or AI.
• Deep knowledge of memory subsystems: HBM, DDR/LPDDR/GDDR, controller architecture, cache coherency and memory consistency models.
• Ownership of a data movement architecture — on-die interconnect, DMA/copy engines, buffer hierarchies, and knowledge of costs for NVLink, UALink, RDMA, ARM CHI, UCIe, PCIe.
• Strong performance modeling and system-level simulation; command of latency, throughput, utilization, and scalability.

Responsibilities

• Define end-to-end memory and data transport architecture for accelerators — from memory technology through the controller, on-die network, and DMA engines to scratchpads.
• Own memory architecture roadmap and energy-efficient interconnect design across multiple silicon generations.
• Perform modeling, simulation, and silicon realization with quantified confidence.

Skills

MSEE/MSCE/PhD

18+ yrs ASIC/SoC

Memory systems expertise

Data movement architecture

Education

MSEE/MSCE or PhD in EE/CE/CS

Job description

USA-CA San Jose Innovation Drive

Full time

R027620

Memory Subsystems and Low-Power Data Transport

Level: Principal Architect/Distinguished Architect

Function: Silicon Architecture — AI/ML Accelerators

The Role

You will define the end-to-end memory and data transport architecture for our accelerators — from the memory technology we bet on, through the controller, on-die network, and DMA engines, all the way into the scratchpads feeding the compute blocks. This is not a "pick the DRAM generation" role. It is ownership of the whole data path across multiple silicon generations, including the performance-versus-power analysis that decides major trade-offs and the co-architecture with internal teams and external memory/IP vendors.

What You'll Own

Memory architecture and roadmap. Multi-generation memory strategy across the portfolio — HBM, LPDDR/GDDR/DDR, on-die SRAM hierarchies and scratchpads. Capacity, bandwidth, and latency requirements derived from real AI workloads (model size, KV-cache growth, prefill versus decode, MoE sparsity). Memory controller and cache/scratchpad microarchitecture: scheduling, address mapping, QoS across compute, DMA, and host traffic.

Data transport and energy-efficient movement. The complete path from memory to the MACs: on-die interconnect (topology, flow control, deadlock avoidance, QoS under mixed traffic), and the DMA, gather/scatter, layout-transform, and multicast engines that make it usable. You own the pJ/bit budget as a first-class constraint.

Modeling and silicon realization. Analytical and simulation-based models that make architectural calls with quantified confidence.

Minimum Qualifications

• MSEE/MSCE or PhD in EE, CE, CS, or related field, or equivalent demonstrated architectural impact.
• 18+ years in ASIC/SoC architecture focused on memory systems and data movement for HPC or AI. Deep knowledge of memory subsystems: HBM, DDR/LPDDR/GDDR, controller architecture, cache coherency and memory consistency models, tiered and pooled memory.
• Demonstrated ownership of a data movement architecture — on-die interconnect, DMA/copy engines, buffer hierarchies, and the dataflow connecting them to compute — including what each protocol (NVLink, UALink, RDMA, ARM CHI, UCIe, PCIe) actually costs in latency, area, and power.
• Strong performance modeling and system-level simulation background; first-principles command of latency, throughput, utilization, and scalability.

Compensation and Benefits

The annual base salary range for this position is USD 200,000.00 To USD 320,000.00

As a valued member of our team, you'll be eligible for a discretionary annual bonus and the opportunity to receive not only a competitive new hire equity grant, but also annual equity awards, connecting your success directly to the company's growth. All subject to relevant plan documents and award agreements.

Broadcom offers a competitive and comprehensive benefits package: Medical, dental and vision plans, 401(K) participation including company matching, Employee Stock Purchase Program (ESPP), Employee Assistance Program (EAP), company paid holidays, paid sick leave and vacation time. The company follows all applicable laws for Paid Family Leave and other leaves of absence.

Broadcom is proud to be an equal opportunity employer. We will consider qualified applicants without regard to race, color, creed, religion, sex, sexual orientation, national origin, citizenship, disability status, medical condition, pregnancy, protected veteran status or any other characteristic protected by federal, state, or local law. We will also consider qualified applicants with arrest and conviction records consistent with local law.

Welcome! Thank you for your interest in Broadcom!

We are a global technology leader that designs, develops and supplies a broad range of semiconductor and infrastructure software solutions.

For more information please visit our video library (http://www.broadcom.com/videos) and check out our Connected by Broadcom (https://www.broadcom.com/support/resources/video-webinar-library?channel=5c9658f7c1d34eca91c6cc16178e5b1b&video=7d72b04ea50046a394e682fab640ce20) series.

Follow us on Linked In Broadcom Inc (https://www.linkedin.com/company/broadcom) .