Accelerated Computing Solutions Analyst - Early… at Marvell… · Santa…

Summary: Marvell Technology is a semiconductor solutions company that provides essential building blocks for data infrastructure. They are seeking an innovative AI Solutions Analyst to research and build co-optimized memory and storage solution architectures for data-centric workloads. Responsibilities: - Develop Next-Gen Solutions: Partner with hardware and software architects, technologists, and IC engineering teams to develop scalable, high-performance, cutting-edge AI/ML infrastructure - Foster Innovation: Spearhead advancements in accelerating AI/ML applications in heterogenous and disaggregated compute architectures, networking, storage, and security domains, focusing on emerging technologies such as CXL, UAL, UEC, low-latency transport protocols, and Post-Quantum Cryptography - Architect and Execute: Lead system architectural design and implement proof-of-concept solutions for diverse workloads, optimizing execution across distributed, heterogeneous memory, compute, and storage environments. Present innovative ideas and substantiated proof-points to senior management and CTOs - Represent in Standards Bodies: Advocate Marvell’s leadership in industry standards bodies and working groups, shaping new standards and emerging technologies aligned with Marvell's strategic interests - Analyze Performance: Conduct advanced characterization and performance analysis of complex workloads, monitoring critical metrics such as response time, latency, and resource utilization - Build Simulation Environments: Create and refine simulation environments for large-scale setups and configurations, especially in the AI/ML domain - Document and Visualize Insights: Develop comprehensive documentation and data visualizations to support analysis and insights, leveraging tools like Python and related libraries - Stay at the Forefront: Keep abreast of the latest developments in generative AI and hardware data analysis, and propose innovative approaches to enhance device tuning and interoperability performance Required Qualifications: - PhD or Master's degree in computer science and engineering, electrical engineering, or related field - HW/SW co-design experience with proven expertise with GPUs, DPUs, FPGAs, or custom hardware accelerators - Processor micro-architectures (preferably ARM), scalar and vector processing pipelines, techniques to optimize memory bound applications - Deep understanding of memory or storage subsystems, disaggregated memory for accelerators, cluster level shared memory approaches, NVMeoF - Familiarity with coherent / non-coherent interconnects, like CXL, UAL, NVLink, and multi-GPU / multi-node communication mechanisms and optimization techniques like SHARP and NCCL - Proficiency in performance benchmarking of AI data processing workloads, system profiling and analyzing end-to-end flows - Experience with simulators & modeling, GPU/CPU cooperative accelerated computing optimizing / accelerating data-centric AI workloads Required Skills: HW/SW co-design, GPU, DPU, FPGA, Custom hardware accelerators, Processor micro-architectures, ARM architecture, Scalar processing pipelines, Vector processing pipelines, Memory bound application optimization, Memory subsystems, Storage subsystems, Disaggregated memory, Cluster level shared memory, NVMeoF, Coherent interconnects, Non-coherent interconnects, CXL, UAL, NVLink, Multi-GPU communication, Multi-node communication, SHARP optimization, NCCL optimization, Performance benchmarking, AI data processing workloads, System profiling, End-to-end flow analysis, Simulation, Modeling, GPU/CPU cooperative accelerated computing, Data-centric AI workloads, Python Benefits: Employee stock purchase plan with a 2-year look back, Family support programs to help balance work and home life, Robust mental health resources to prioritize emotional well-being, Recognition and service awards to celebrate contributions and milestones