About the job
At d-Matrix, we are pioneering the potential of generative AI to revolutionize technology. Positioned at the leading edge of software and hardware innovation, we are dedicated to pushing the limits of what is achievable. Our workplace culture is built on mutual respect and collaboration.
We embrace humility and value direct communication. Our team is diverse and inclusive, allowing for a variety of perspectives that lead to superior solutions. We are looking for individuals who are eager to take on challenges and are motivated by results. Are you ready to discover your playground? Together, we can explore the infinite possibilities of AI.
This position requires you to work on-site at our Santa Clara, CA headquarters for three days a week in a Hybrid work model.
Role Overview: Principal Architect - Performance Analysis and Modeling
d-Matrix is searching for exceptional computer architects to enhance AI application performance at the convergence of hardware and software, with a keen focus on cutting-edge hardware technologies (such as DIMC, D2D, 3D-DRAM, etc.) and innovative workloads (like generative inference, etc.). Our acceleration philosophy spans the entire system, encompassing efficient tensor cores, storage, and data movements, along with co-designing dataflow and collective communication techniques.
Key Responsibilities:
Analyze the latest machine learning workloads (including multi-modal LLMs, CoT reasoning models, and video/audio generation).
Contribute to the Hardware and Software features that empower the next generation of inference accelerators in data centers.
Stay current with research in ML Architecture and Algorithms, collaborating with various partner teams including Product, Hardware Design, Compiler, Inference Server, and Kernels.
Daily tasks will involve (1) analyzing the properties of emerging machine learning algorithms and workloads to discern functional and performance implications, (2) creating analytical models to forecast performance across current and future generations of d-Matrix hardware, and (3) proposing new hardware/software features to facilitate or accelerate these algorithms.
Qualifications:
Minimum:
Bachelor's degree in Electrical Engineering or related field with over 10 years of industry experience, or a Master's degree preferred with at least 8 years of experience.
Comprehensive understanding of relevant areas through academic or industry experience, including computer architecture, hardware-software co-design, and performance analysis.
