Baidu USA
SoC Memory Subsystem Architect
Sunnyvale,CA +1 · Full-time · Posted Jun 17, 2021
- Location
- Sunnyvale,CA
Sunnyvale, CA, United States - Experience
- 10+ years.
- Education
- Master or PhD in Electrical or Computer Engineering.
About the job
Leads architecture and development of a unified memory subsystem for Baidu’s AI accelerator SoCs and distributed AI training systems.
What you'll do
- Develop memory-system architecture and microarchitecture, evaluating SoC fabrics, caches, coherency, and MMUs.
- Collaborate with software, SoC design, and hardware teams to define, build, and test a unified memory system.
- Analyze workloads and simulation results, optimize performance, and debug functional or performance issues from models through silicon.
What they ask for
- 10+ years of experience in silicon architecture or IP design focused mainly on memory subsystems.
- Strong understanding of distributed AI training systems’ memory-subsystem requirements; experience in digital hardware design and system performance analysis/debug.
- Solid background in system interconnects, System MMUs, caches, and memory technologies.
- Master or PhD in Electrical or Computer Engineering; excellent communication skills in English and Chinese.
Nice to have
- Familiarity with software and operating-system memory practices and requirements.
- Experience analyzing data with Excel, Perl, or Python.
Tools
Excel, Perl, Python, HBM, GDDR, DDR, LPDDR4/5, RTL
AI focus
Architects a unified memory subsystem for AI accelerator SoCs and distributed AI training systems.
Model-written summary · the original posting is authoritative