Principal AI Memory Architect

Samsung Semiconductor Inc.; San Jose (CA); Hybrid; USD 219,000 - 351,000; Full time

Benefits offered by this job

4+ weeks paid time off

Charitable giving match

Fertility/adoption support

Onsite cafe and gym

Job summary

Samsung Semiconductor Inc. is seeking a Hands-on Principal Engineer to own full-stack AI memory solutions for large-scale inference.

You will design tiered memory hierarchies, optimize MoE and transformer workloads, and drive performance in production stacks such as vLLM, HiCache, and TensorRT-LLM. You will influence memory-product requirements, architecture direction, and roadmap decisions while collaborating with customers and internal teams to deliver scalable, cost-efficient AI memory

Qualifications

• BS in Computer/Electrical/Electronic Engineering or Computer Science; 10+ years in systems engineering.
• MS in Computer/Electrical/Electronic Engineering or Computer Science preferred.
• 10+ years in large-scale LLM inference or GPU systems performance.

Responsibilities

• Define and own memory-stack requirements for inference at fleet scale.
• Lead architecture reviews and mentor engineers across memory/IO layers.
• Collaborate with customers and partners at CTO level on memory products.

Skills

Technical leadership

AI memory systems

Memory bandwidth

Large-scale inference

Performance engineering

Education

BS in Computer/Electrical/Electronic Engineering or CS

MS in Computer/Electrical Engineering or CS

Tools

vLLM

SGLang HiCache

TensorRT-LLM

llama.cpp

Job description

Samsung Semiconductor Inc. is seeking a Hands-on Principal Engineer to own full-stack AI memory solutions for large-scale inference.

You will design tiered memory hierarchies, optimize MoE and transformer workloads, and drive performance in production stacks such as vLLM, HiCache, and TensorRT-LLM. You will influence memory-product requirements, architecture direction, and roadmap decisions while collaborating with customers and internal teams to deliver scalable, cost-efficient AI memory