Lead, End-to-End Inference Performance

SambaNovaSystems; San Jose (CA); On-site; USD 245,000 - 325,000; Full time

Job summary

SambaNova Systems is seeking an Architect for the Inference Systems Performance team to own end-to-end performance of large-scale LLM inference. You will study how requests move from tokenization to decode and how a deployment is sized to meet customer SLOs.

This role spans workload capture, benchmarking, and performance modeling to guide capacity planning and system design. You'll define strategies, build benchmarking capabilities, and develop tooling to profile a distributed inference

Qualifications

• 12+ years of experience in performance engineering
• Deep expertise in end-to-end performance analysis of distributed systems
• Proven command of workload generation and simulation
• Demonstrated ability to enter unfamiliar domains and apply core performance methods
• Ability to lead cross-functional efforts and mentor senior engineers
• Experience representing an organization credibly to customers and partners
• Track record of scoping and delivering high-complexity work with impact

Responsibilities

• Define and drive the technical strategy for inference-systems performance including workload capture, benchmarking, modeling, and simulation
• Build workload-capture and agentic-benchmarking capability to faithfully reflect production traffic
• Own the performance-modeling and simulation practice to predict how configurations affect output and planning
• Drive end-to-end profiling tooling to produce actionable profiles of a distributed inference pipeline
• Serve as the senior technical voice across model-optimization, systems, hardware, and product
• Represent SambaNova's performance story to customers and partners
• Mentor and multiply capability by raising engineers' skills and patterns
• Drive resolution of novel, ambiguous challenges spanning organizational boundaries

Job description

SambaNova Systems is seeking an Architect for the Inference Systems Performance team to own end-to-end performance of large-scale LLM inference. You will study how requests move from tokenization to decode and how a deployment is sized to meet customer SLOs.

This role spans workload capture, benchmarking, and performance modeling to guide capacity planning and system design. You'll define strategies, build benchmarking capabilities, and develop tooling to profile a distributed inference