Job Summary
In this AI/ML HBF ASIC Performance Modeling position, you will develop high-performance simulation models and modeling methodologies for Sandisk’s next-generation High Bandwidth Flash (HBF) AI Storage Solutions. You will work at the intersection of ASIC architecture, AI/ML workloads, memory systems, and system-level performance engineering to evaluate architectural choices before silicon implementation. You will develop and enhance SystemC/TLM-based performance models representing key HBF ASIC components and their interactions with GPU/TPU/xPU accelerators. The role requires strong SystemC/C++ development skills combined with a solid understanding of computer architecture and performance analysis.
Key Responsibilities
- Develop, maintain, and enhance SystemC/TLM 2.0 transaction-level performance models for HBF ASICs and AI Storage Solutions, enabling rapid architectural exploration before RTL implementation
- Model critical ASIC subsystems including UCIe/PCIe/CXL interfaces, NoC, DMA engines, queues and buffers, memory controllers, embedded processors, NAND interfaces, and high-bandwidth datapaths
- Develop configurable modeling infrastructure allowing architects to perform architectural what-if analysis, parameter sweeps, bottleneck identification, and design-space exploration
- Characterize AI/ML and LLM workloads, including training and inference workloads, Dense and Mixture-of-Experts models, attention/KV-cache traffic, and accelerator-to-storage data movement
- Analyze end-to-end performance between GPU/TPU/xPU accelerators and HBF, identifying opportunities to improve effective bandwidth, latency, concurrency, and hardware utilization
- Partner with ASIC architects to evaluate alternative microarchitectures and quantitatively influence architecture specifications and product decisions
- Develop performance metrics, statistics, traces, visualization, regression, and analysis methodologies that enable rapid identification and root-cause analysis of performance bottlenecks
- Correlate SystemC model behavior against RTL simulation, emulation, FPGA prototypes, and silicon measurements where available, improving model fidelity throughout the product lifecycle and Model-to-RTL/hardware validation.
Required Qualifications
- Bachelors/Masters/Phd degree with 5+ years of relevant experience in Computer Engineering, Electrical Engineering, or Computer Science
- Strong hands-on expertise developing simulation and performance models using SystemC, TLM 2.0, C/C++, and object-oriented software design
- Strong understanding of computer architecture, including NoC/interconnects, memory hierarchy, queues, arbitration, DMA, caching, memory controllers, and high-bandwidth I/O
- Accurately model architecture-level timing, concurrency, bandwidth, latency, contention, backpressure, arbitration, resource utilization, and pipeline behavior
- Experience analyzing system performance including throughput, latency, contention, utilization, queue occupancy, backpressure, and bottleneck identification
- Experience developing performance models for AI/ML accelerators, GPUs/TPUs/NPUs, datacenter ASICs, storage controllers, or memory systems
- Knowledge of LLM training and inference, including attention, KV cache, Mixture-of-Experts, batching, and memory-bandwidth requirements
- Experience with SystemC/TLM model integration, RTL/TLM co-simulation, model correlation, regression infrastructure, and architectural design-space exploration.