AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix
Netflix uses AI agents to automate performance profiling and cut infrastructure costs at scale
“agent doesn't know specific details about your platforms and your frameworks uh and your internal code base patterns”
Netflix staff engineer Rajat Shah presents a practitioner playbook for using AI agents in performance engineering workflows to improve throughput and reduce compute costs. The talk identifies a compounding problem: coding agents ship code 10x faster but produce non-performant code ignorant of internal platform patterns, causing infrastructure costs to rise proportionally. Netflix's solution applies AI agents to automate the profiling and bottleneck-identification work traditionally done by human performance engineers.