Senior NPU Architect for Next-Gen SoCs

Samsungresearchamerica; Mountain View (CA); On-site; USD 204,000 - 306,000; Full time

Job summary

Samsung Research America seeks an accomplished NPU Architect to lead architectural and micro-architectural work for next‑gen SoCs. You will drive exploration, performance modeling, and HW/SW co-design while shaping power and area efficiency for Galaxy devices.

The role requires deep expertise in CNN/LLM models, Python/C++, and ML frameworks, with leadership across hardware and software teams in a high‑visibility setting.

Qualifications

• Bachelor’s, Masters or PhD in Electrical, Computer Science, or related field.
• 14+ years of experience in NPU architecture & development.
• Hands-on experience in cycle-accurate performance and functional models.
• Strong knowledge of CNN and LLM models.
• Proficient in Python, System C and/or C++.
• Experience with ML frameworks (TensorFlow, PyTorch, ONNX).
• Experience in SoC micro-architecture and interface subsystems.
• Self-motivated, team-oriented problem solver.

Responsibilities

• Propose NPU architectural and micro-architectural features to boost PPA.
• Perform high-level performance modeling and analysis of NPU features.
• Develop cycle-accurate performance models and correlate with silicon.
• Deliver architecture proposals to design teams and leadership.
• Analyze benchmarks and workloads for architecture optimization.
• Map AI use cases to NPU and optimize HW/SW co-design.
• Collaborate across SOC, Driver, OS, and System teams with documentation.

Skills

NPU architecture

Cycle-accurate modeling

Python

C++

System C

CNN/LLM models

ML frameworks

SoC micro-architecture

Hardware-Software co-design

Team collaboration

Education

Bachelor’s/Master’s/PhD in Electrical or Computer Science

Tools

TensorFlow

PyTorch

ONNX

Python

C++

System C

Job description

Samsung Research America seeks an accomplished NPU Architect to lead architectural and micro-architectural work for next‑gen SoCs. You will drive exploration, performance modeling, and HW/SW co-design while shaping power and area efficiency for Galaxy devices.

The role requires deep expertise in CNN/LLM models, Python/C++, and ML frameworks, with leadership across hardware and software teams in a high‑visibility setting.