Multimodal Inference Engineer

OpenAI; United States; Remote; USD 325,000 - 490,000; Full time

Benefits offered by this job

Medical insurance

Mental health support

401(k) plan with 50% matching

Generous time off

Paid parental leave

Learning & development stipend

Job summary

OpenAI is seeking a software engineer to scale multimodal model inference. You will design and optimize the infrastructure for serving real-time audio, image, and other MM workloads in production, collaborating with researchers and product teams to push capabilities forward.

You will own end-to-end delivery, including GPU utilization, tensor parallelism, and model-parallel workflows, in a fast-moving environment with strong emphasis on performance.

Qualifications

• Experience building inference infrastructure for large-scale multimodal models.
• Experience with GPU-based ML workloads and model parallelism.
• Ability to work across research, product, and infra teams.
• Comfort operating in fast-evolving environments.

Responsibilities

• Design and implement inference infrastructure for large-scale multimodal models.
• Optimize systems for high-throughput, low-latency delivery of image and audio inputs and outputs.
• Enable experimental research workflows to transition into reliable production services.
• Collaborate with researchers, infra teams, and product engineers to deploy capabilities.
• Contribute to system-level improvements including GPU utilization and hardware abstraction.

Skills

Inference systems

GPU-based ML workloads

Distributed compute

High-throughput data handling

vLLM

TensorRT-LLM

Model parallel systems

End-to-end ownership

Tools

vLLM

TensorRT-LLM

Job description

OpenAI is seeking a software engineer to scale multimodal model inference. You will design and optimize the infrastructure for serving real-time audio, image, and other MM workloads in production, collaborating with researchers and product teams to push capabilities forward.

You will own end-to-end delivery, including GPU utilization, tensor parallelism, and model-parallel workflows, in a fast-moving environment with strong emphasis on performance.