Staff Software Engineer, Inference

In this role you will help operate Claude’s inference stack at scale, enabling reliable and high-performance model serving for millions of users. You’ll own end-to-end infrastructure work—from intelligent request routing to fleet-wide orchestration across accelerators—balancing compute efficiency with research acceleration. You’ll work with a cross-functional team to remove blockers, optimize performance, and deploy multi-accelerator deployments. This is a chance to influence scalable AI infrastructure at a company focused on safe, beneficial AI.

• competitive compensation and benefits
• optional equity donation matching
• generous vacation and parental leave
• flexible working hours
• office space for collaboration

• Identify and tackle infrastructure blockers to enable large-scale model serving
• Drive performance optimization across distributed systems and inference pipelines
• Implement and optimize load balancing, request routing, and traffic management
• Support multi-accelerator deployments and cloud infrastructure in AWS/GCP
• Build and maintain production-grade deployment pipelines for new models and features
• Collaborate with research teams to enable high-performance inference for evolving architectures
• Analyze observability data to tune performance in production
• Manage multi-region deployments and geographic routing across global users

• Significant software engineering experience with distributed systems
• Familiarity with performance optimization and large-scale service orchestration
• Experience with LLM inference optimization, batching, and caching (encouraged)
• Experience with load balancing, request routing, or traffic management systems
• Knowledge of Kubernetes and cloud infrastructure (AWS, GCP)
• Proficiency in Python or Rust
• Ability to work across academia and industry with impact-focused mindset
• Interest in machine learning systems and infrastructure
• Care about societal impacts of AI

• results-oriented
• flexibility and impact
• strong collaboration and communication
• LLM inference optimization
• batching strategies
• multi-accelerator deployments