SoC Architect
Job description
SoC Architect at D Matrix.
About the role
You will own the definition and execution of the SoC architecture for d-Matrix's AI compute engine, shaping the silicon strategy from requirements to implementation. You will analyze AI workloads such as transformers and CNNs to uncover bottlenecks and drive architectural innovations across the full stack. This role demands leadership in hardware-software co-design, performance modeling, and cross-functional collaboration to ensure product excellence. You will design, implement, and evaluate efficient deep neural network architectures tailored for the demands of generative AI. Your work will directly influence the direction of our Corsair product family and next-generation initiatives. You will partner closely with RTL and DV engineers to align on performance, power, thermal, and reliability goals. Ultimately, you will simplify complex problems into elegant and efficient solutions that define the leadership position of d-Matrix technology.
Key facts
What you'll do
Lead the design and delivery of SoC requirements and architecture that delivers exceptional customer experiences.
Analyze AI workloads such as transformers & CNNs across existing and new emerging product categories to identify bottlenecks and opportunities for improvement.
Identify areas of improvement and drive the architecture to deliver the product through design, implementation, and evaluation of efficient deep neural network architectures.
Conduct research to guide hardware/software co-design of SIMD engine, control plane, data reshape engine, data convertor blocks, NoC substrate and many other blocks.
Work on various phases of our flagship product (Corsair I and II) and the Next-Generation product ranging from performance projection, performance modeling, performance correlation and performance debug aspects.
Engage in Kernel Development and assist in the Networking Fabric Enhancements.
Develop and maintain tools for high-level simulation and research to enable architectural exploration of current and future generation products.
Engage and collaborate with SW team to meet stack development milestones and ensure alignment across the compute stack.
Port customer workloads, optimize them for deployment, generate reference implementations and evaluate performance on d-Matrix platforms.
Report and present progress timely and effectively to both technical and executive stakeholders.
Contribute to publications of papers and intellectual properties that showcase d-Matrix innovation.
Drive architectural exploration and modeling for memory, interconnects, packaging, and data movement subsystems.
Champion early identification of architectural risks and define mitigation strategies in collaboration with cross-functional teams.
Define metrics and benchmarks that validate architecture quality and customer impact throughout the product lifecycle.
Requirements
MS in Computer Science, Electrical and Computer Engineering, or a related scientific discipline with 5+ year of relevant experience in compute architecture and system design area.
A problem solver, self-starter, be able to effectively multi-task and simplify complex problems to come up with elegant and efficient solution (no compromise here!)
5 or more years of demonstrated experience leading Architecture or Design Development of SoCs.
Expert knowledge of AI SoC Architecture, requirements, and definition.
Be able to identify potential architectural issues and design flaws early in the design phase to provide effective solutions and recommendations.
Experience with CPU, GPU, ASIC and ML accelerator architecture exploration and architectural modeling.
High proficiency in performance modeling, ranging from simple analytical model to complex cycle accurate performance model and correlation.
Deep, wide and current knowledge in machine learning and modern deep learning is required.
Hands-on experience with CNN, RNN, Transformer neural network architectures.
Nice to have
Experience with specialized HW accelerator systems for deep neural network is preferred.
PhD in Computer Science, Electrical and Computer Engineering, or a related scientific discipline with 5+ year of relevant experience in compute architecture and system design area.
High proficiency in algorithm analysis, data structure, and Python programming is required. Proficiency with C/C++ programming is preferred.
Passionate about AI and thriving in a fast-paced and dynamic startup culture.
Practical notes
Location: Hybrid (Santa Clara, CA)
Engagement: Full-time