Performance Engineer, Inference Systems
Core
Optimizing the throughput, latency, reliability, and correctness of Anthropic's large-scale AI inference fleet serving Claude.
Role type
Senior IC performance engineer (inference systems)
Builds
Observability dashboards, correctness evaluation pipelines, and performance modeling tools for the inference stack.
Domain
AI/ML infrastructure, large-scale distributed systems, GPU/TPU acceleration
Deliverable
production ML models
Required skills
performance engineering, profiling, roofline analysis, latency/throughput optimization, root-cause investigation, Python, data analysis (SQL/pandas), correctness evaluation design
Preferred skills
ML systems experience, GPU/TPU performance concepts, reliability engineering for high-throughput services, model evaluation pipelines, observability for distributed systems
Technologies
Python, SQL, pandas, GPU/TPU, accelerator kernels, model servers, distributed routing, autoscaling
Responsibilities
Run cross-layer performance investigations to size gaps between actual and theoretical performance; own and improve correctness evaluation pipelines; build observability and modeling tools; partner with kernel and serving teams to land optimizations; prioritize opportunities by impact and effort.
Seniority
Senior, hands-on IC
