--- title: AI Agents for Performance: Ship Faster, Pay Less — Rajat... description: A Netflix staff engineer shares a practitioner's playbook for using AI agents to automate performance engineering. The core problem: AI coding agents ship code... canonical: https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix | daily.dev og:description: A Netflix staff engineer shares a practitioner's playbook for using AI agents to automate performance engineering. The core problem: AI coding agents ship code... og:url: https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr og:image: https://api.daily.dev/og/posts/v8bFqtdmr.png og:image:alt: AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 33 min read · 1 upvotes · 0 comments ## Summary A Netflix staff engineer shares a practitioner's playbook for using AI agents to automate performance engineering. The core problem: AI coding agents ship code 10x faster but often produce suboptimal code, driving up compute costs. The proposed solution uses LLM agents to read profiling data, identify performance anti-patterns (like O(N²) loops or unnecessary object allocations), clone the relevant Git repo, and automatically generate code review PRs — all in under 5 minutes. Key components include a centralized Git-based catalog of patterns and anti-patterns that acts as long-term memory for stateless LLM agents, canary deployments for ground-truth validation before human review, and a shift-left strategy that moves performance checks from reactive (post-production) to proactive (code authoring phase). The talk recommends a phased automation approach: start with LLM-assisted identification (level 1), progress to fully orchestrated agentic workflows with tool integrations (level 2), and only pursue fully autonomous agents (level 3) with heavy investment in security guardrails and sandboxing. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can an AI agent use CPU profiling data to find performance bugs automatically? An AI agent can read structured profiler output (call stack, self CPU, inclusive CPU per method) directly, since profiler formats look similar regardless of the underlying language such as Java, Python, or Go. The agent matches call stack patterns to known anti-patterns like O(N squared) loops or loop invariants, then checks out the exact production Git commit to locate and fix the offending code, producing a code review in under 5 minutes in one real example. _Engineers building agentic performance tooling can find deeper technical breakdowns like this one on daily.dev._ ### What is a pattern catalog for AI coding agents and why use markdown files instead of a vector database? A pattern catalog is a centralized, human- and agent-readable Git repository of confirmed code patterns and anti-patterns, built incrementally from real profiling findings across services. Markdown files work because they let both humans and coding agents author and query entries directly without vector search overhead, and confidence scores per entry grow as more services confirm the same anti-pattern, making future agent suggestions more trustworthy. _Teams designing shared AI-agent memory systems can track approaches like this via daily.dev._ ### Why should AI agents send performance fixes as code reviews instead of pushing directly to production? Human approval stays required because modifying code already running correctly in production to optimize it is risky when business context or test coverage is incomplete. Before requesting human review, the agent should run integration and unit tests to verify functional correctness, then validate the fix with a canary deployment comparing CPU and latency between old and new code over roughly 10 minutes, escalating only if error rates stay flat. _Developers weighing autonomy levels for AI agents in production workflows can follow this reasoning on daily.dev._ ## Similar posts on daily.dev - [When code is abundant](https://daily.dev/posts/when-code-is-abundant-gezbqtizx) · GitLab · 0 upvotes · 0 comments - [Agentic Code Review](https://daily.dev/posts/agentic-code-review-0ue8fcsa1) · Addy Osmani · 147 upvotes · 14 comments - [A grounded take on agentic coding for production environments](https://daily.dev/posts/a-grounded-take-on-agentic-coding-for-production-environments-eqmzfdloq) · Ivan Velichko · 35 upvotes · 1 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents) [View this post on daily.dev](https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix","url":"https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr"},"datePublished":"2026-07-28T01:14:17.539Z","dateModified":"2026-09-14T07:15:05.350Z","description":"A Netflix staff engineer shares a practitioner's playbook for using AI agents to automate performance engineering. The core problem: AI coding agents ship code...","image":"https://i.ytimg.com/vi/CgsWxRUY5Eo/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/CgsWxRUY5Eo/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents","timeRequired":"PT33M","video":{"@type":"VideoObject","name":"AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix","description":"A Netflix staff engineer shares a practitioner's playbook for using AI agents to automate performance engineering. The core problem: AI coding agents ship code...","thumbnailUrl":"https://i.ytimg.com/vi/CgsWxRUY5Eo/sddefault.jpg","uploadDate":"2026-07-28T01:14:17.539Z","duration":"PT33M","url":"https://api.daily.dev/r/v8bFqtdmr","embedUrl":"https://www.youtube.com/embed/CgsWxRUY5Eo"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"AI Agents for Performance: Ship Faster, Pay Less — Rajat Shah, Netflix"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-agents-for-performance-ship-faster-pay-less-rajat-shah-netflix-v8bfqtdmr#faq","mainEntity":[{"@type":"Question","name":"How can an AI agent use CPU profiling data to find performance bugs automatically?","acceptedAnswer":{"@type":"Answer","text":"An AI agent can read structured profiler output (call stack, self CPU, inclusive CPU per method) directly, since profiler formats look similar regardless of the underlying language such as Java, Python, or Go. The agent matches call stack patterns to known anti-patterns like O(N squared) loops or loop invariants, then checks out the exact production Git commit to locate and fix the offending code, producing a code review in under 5 minutes in one real example. Engineers building agentic performance tooling can find deeper technical breakdowns like this one on daily.dev."}},{"@type":"Question","name":"What is a pattern catalog for AI coding agents and why use markdown files instead of a vector database?","acceptedAnswer":{"@type":"Answer","text":"A pattern catalog is a centralized, human- and agent-readable Git repository of confirmed code patterns and anti-patterns, built incrementally from real profiling findings across services. Markdown files work because they let both humans and coding agents author and query entries directly without vector search overhead, and confidence scores per entry grow as more services confirm the same anti-pattern, making future agent suggestions more trustworthy. Teams designing shared AI-agent memory systems can track approaches like this via daily.dev."}},{"@type":"Question","name":"Why should AI agents send performance fixes as code reviews instead of pushing directly to production?","acceptedAnswer":{"@type":"Answer","text":"Human approval stays required because modifying code already running correctly in production to optimize it is risky when business context or test coverage is incomplete. Before requesting human review, the agent should run integration and unit tests to verify functional correctness, then validate the fix with a canary deployment comparing CPU and latency between old and new code over roughly 10 minutes, escalating only if error rates stay flat. Developers weighing autonomy levels for AI agents in production workflows can follow this reasoning on daily.dev."}}]} ```