Hardware / Software CoDesign Engineer - 3P

About this role

OpenAI seeks a Hardware/Software CoDesign Engineer to optimize AI accelerators and silicon for large-scale machine learning workloads. You'll collaborate with hardware vendors, ML engineers, and compiler teams to co-design next-generation hardware tailored for efficient training and inference of advanced AI models.

What you'll do

• Co-design hardware for programmability and performance with external hardware vendors
• Develop performance models and estimates for critical kernels across different hardware configurations
• Build system-level performance models to optimize compute, memory hierarchy, and networking decisions
• Partner with ML, kernel, and compiler engineers to translate their requirements into hardware features
• Evaluate and influence hardware partner roadmaps to align with OpenAI's datacenter and workload needs
• Manage coordination between internal teams and external partners on hardware optimization initiatives

What they're looking for

• Software/hardware co-design
• GPU and AI accelerator architecture
• CUDA, Triton, or similar accelerator programming languages
• System performance modeling and analysis
• C/C++ and Python programming
• Deep learning fundamentals and chip architecture
• Low-precision ML optimization
• Collaborative cross-functional communication

Benefits

• Hybrid work model: 3 days per week in San Francisco office
• Relocation assistance provided for new employees