Director, AI Workload Intelligence & Memory Systems

SK hynix America; San Jose (CA); On-site; USD 250,000 - 300,000; Full time

Benefits offered by this job

Top-tier health insurance

Paid time off

401k matching

FSA for Health Care & Dependent Care

Educational reimbursement up to $10,00

Donation matching & volunteering

Free meals

Job summary

SK hynix America seeks a Director of AI Workload Intelligence to lead global AI workloads analysis for HBF applicability and data-placement criteria. You will guide memory system strategies and collaborate with Compiler/Runtime teams from research to productization.

The role requires hands-on experience in AI inference systems, knowledge of Transformers/LLMs, and extensive tooling expertise across PyTorch, Hugging Face, and ONNX Runtime.

Qualifications

• Proven experience leading global AI system architecture initiatives and cross-functional teams.
• Hands-on experience in AI inference systems, ML systems, LLM serving, or heterogeneous accelerator software.
• Strong understanding of Transformers, LLMs, MoE, embedding/retrieval workloads, and multimodal inference.
• Experience with PyTorch, Hugging Face, vLLM, SGLang, TensorRT-LLM, ONNX Runtime, Triton Inference Server.
• Experience in latency, throughput, memory footprint, and data movement analysis.

Responsibilities

• Lead global collaboration and drive AI workload-based HBF technology development from research to PoC and productization.
• Analyze AI models, algorithms, and workload trends for HBF applicability.
• Classify HBF-relevant memory objects such as weights, KV cache, activations, embeddings, adapters, and MoE experts.
• Define model-level criteria for data placement, caching, prefetching, and offload decisions.
• Analyze HBF integration points in vLLM, SGLang, TensorRT-LLM, PyTorch, and ONNX Runtime.
• Identify HBF-applicable AI model families and inference data objects.
• Develop memory-use taxonomy and HBF applicability criteria.
• Provide model-driven inputs to Compiler/Runtime and AI Inference SW Stack teams.

Skills

AI system architecture

Cross-functional leadership

LLM model analysis

AI inference systems

Transformer knowledge

Tools

PyTorch

Hugging Face

vLLM

SGLang

TensorRT-LLM

ONNX Runtime

Triton Inference Server

Job description

SK hynix America seeks a Director of AI Workload Intelligence to lead global AI workloads analysis for HBF applicability and data-placement criteria. You will guide memory system strategies and collaborate with Compiler/Runtime teams from research to productization.

The role requires hands-on experience in AI inference systems, knowledge of Transformers/LLMs, and extensive tooling expertise across PyTorch, Hugging Face, and ONNX Runtime.