Software Engineer – ML Infrastructure

People Culture Talent; San Francisco (CA); On-site; USD 190,000 - 230,000; Full time

Job summary

People Culture Talent seeks an experienced ML Infrastructure Software Engineer to design, build, and operate infrastructure for ML workloads. You’ll partner with researchers and engineers to enable experimentation, training, evaluation, and production deployment with a focus on reliability and performance.

Responsibilities include creating robust pipelines, improving compute utilization, and leading architectural decisions from design to deployment.

Qualifications

• Senior-level software engineer with ownership of ML infrastructure.
• Proven design, build and operation of core ML infra.
• Hands-on PyTorch, model serving engines, and distributed training.
• Experience with distributed systems and GPU workloads.
• Background in RL post-training infra, RLHF, or automated model evaluation.
• Strong systems design, debugging and performance optimization.
• Clear communication of architectural trade-offs and decisions.
• Experience maintaining and evolving systems over time.

Responsibilities

• Design and build infrastructure for model training and production serving.
• Develop reliable pipelines and shared tools for ML workloads.
• Improve compute utilization, scheduling and performance.
• Use profiling and monitoring to identify bottlenecks and troubleshoot.
• Automate testing and measurement to assess changes.
• Lead architectural decisions through design, deployment and operation.
• Collaborate with researchers and engineers to translate requirements into systems.

Skills

Senior ownership

Distributed systems

System design & performance

Communication

End-to-end ownership

Tools

PyTorch

Model serving engines

Distributed training frameworks

GPU compute experience

Job description

The Role

We’re seeking an experienced ML Infrastructure Software Engineer to build and improve infrastructure for machine learning workloads. You’ll partner with researchers and engineers to support experimentation, training, evaluation, and production deployment, with a focus on reliability and performance.

What You’ll Own

• Design and build infrastructure for model training and production serving.

• Develop reliable pipelines and shared tools that help teams run and manage machine learning workloads.

• Improve compute utilization, workload scheduling, and performance across distributed systems.

• Use profiling and monitoring to identify bottlenecks, troubleshoot failures, and guide improvements.

• Automate testing and measurement to help teams assess changes consistently.

• Lead architectural decisions, explain trade-offs, and carry projects from initial design through deployment and ongoing operation.

• Collaborate with researchers and engineering teams to turn evolving requirements into maintainable systems.

What You’ll Bring

• Substantial software engineering experience with technical ownership at Senior Engineer level or equivalent.

• A track record of personally designing, building, and operating core machine learning infrastructure.

• Hands-on experience with PyTorch, model serving engines, and distributed training frameworks

• Strong experience with distributed systems and large-scale compute environments, including GPU-based workloads.

• Solid background in reinforcement learning (RL) post-training infrastructure, RLHF, or automated model evaluation pipelines

• Practical experience with modern machine learning frameworks and their supporting infrastructure.

• Strong systems design, debugging, and performance optimization skills.

• Clear, concise communication and the ability to explain your contributions, architectural choices, and technical trade-offs.

• Experience owning systems over time, including maintenance, reliability, and improvements as requirements evolve.

Preferred experience

• Experience supporting both training infrastructure and production model deployment.

• Experience with large language model workloads is a plus, but not required. Infrastructure experience with other types of machine learning models is equally relevant.