Principal Architect, Performance Analysis and Modeling
d-MatrixJob Title
Principal Architect, Performance Analysis and Modeling
Role Summary
The Principal Architect will analyze emerging ML workloads, build analytical performance models, and propose hardware and software features to optimize inference accelerator performance. The role sits at the intersection of hardware and software and partners with Product, Hardware Design, Compiler, Inference Server, and Kernel teams.
Work arrangement: hybrid (onsite at Santa Clara, CA headquarters three days per week); remote candidates within the United States will be considered.
Experience Level
Senior — typically requires 8–10+ years of relevant industry experience in computer architecture, hardware-software co-design, or performance modeling.
Responsibilities
Key responsibilities include workload analysis, modeling, and cross-team collaboration to drive accelerator features.
- Analyze properties and performance characteristics of emerging ML workloads (multi-modal LLMs, chain-of-thought reasoning, generative video/audio models).
- Create analytical performance models to project performance on current and future d-Matrix hardware.
- Develop, validate, or use architecture simulators and performance analysis methodologies.
- Propose HW/SW co-design features (dataflow, tensor core optimizations, storage and data movement, collective communication) to accelerate inference.
- Collaborate with Product, Hardware Design, Compiler, Inference Server, and Kernel teams to specify and prioritize features.
- Monitor and incorporate the latest ML architecture and algorithm research into product work.
Requirements
Must-have technical skills and desirable additions.
- Must have: Strong background in computer architecture, hardware-software co-design, and performance modeling for ML accelerators.
- Proficiency in C/C++ and/or Python for prototyping and analysis.
- Experience developing analytical performance models and working with architecture simulators for performance analysis.
- Practical knowledge of DNN fundamentals and inference performance trade-offs.
- Self-motivated team player with strong collaboration and initiative.
- Nice to have: Research publication record in top-tier architecture or ML venues (ISCA, MICRO, ASPLOS, HPCA, DAC, MLSys).
Education Requirements
Minimum: Bachelor of Science in Electrical Engineering (BSEE) with 10+ years of industry experience. Preferred: Master of Science in Electrical Engineering (MSEE) with ~8+ years of industry experience. Degrees in Electrical Engineering, Computer Engineering, Computer Science, or related technical fields are relevant.
About the Company
Company: d-Matrix
Headquarters: Santa Clara, California, United States
d-Matrix is a Santa Clara–based startup developing highly programmable in-memory computing architectures and accompanying software to accelerate generative AI and other AI workloads, focusing on hardware-software co-design for cloud and edge applications.
