GPU & Accelerator Architecture Lead for Next‑Gen Silicon

OpenAI; San Francisco (CA); Hybrid; USD 180,000 - 280,000; Full time

Benefits offered by this job

Relocation assistance

Job summary

OpenAI’s Hardware organization seeks a 3P Hardware Architecture Expert to guide GPU/accelerator decisions for AI infrastructure. You will evaluate compute, memory, and interconnect tradeoffs, translate vendor specs into workload impact, and lead early ES silicon bring-up with partners like NVIDIA and AMD.

Based in San Francisco with a hybrid model (3 days in the office) and relocation assistance, you will drive architecture choices and validate performance for next-generation AI workloads.

Qualifications

• Deep expertise in GPU or accelerator architecture.
• Comfortable engaging with silicon vendors at a technical level.
• Hands-on benchmarking, profiling, and performance analysis.
• Ability translate hardware details into system/workload impact.

Responsibilities

• Engage deeply with silicon vendors (e.g NVIDIA & AMD) on GPU and accelerator architecture tradeoffs.
• Analyze performance, power, and efficiency characteristics of next-generation hardware.
• Translate vendor specifications into expected real-world performance for AI workloads.
• Evaluate compute throughput, memory systems, data types/precision, and interconnect scaling.
• Run benchmarks and profiling to validate hardware performance against workload requirements.
• Lead early bring-up and evaluation of engineering sample silicon.
• Partner with performance modeling and system architecture teams to align measured vs. modeled behavior.
• Provide actionable feedback to vendors to influence future silicon design and roadmap decisions.

Skills

GPU architecture

Accelerator architecture

Vendor engagement

Benchmarking

Job description

OpenAI’s Hardware organization seeks a 3P Hardware Architecture Expert to guide GPU/accelerator decisions for AI infrastructure. You will evaluate compute, memory, and interconnect tradeoffs, translate vendor specs into workload impact, and lead early ES silicon bring-up with partners like NVIDIA and AMD.

Based in San Francisco with a hybrid model (3 days in the office) and relocation assistance, you will drive architecture choices and validate performance for next-generation AI workloads.