<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Anand Kumar — Writing</title><description>Notes on AI infrastructure, inference, and operations.</description><link>https://anand-kumar.dev/</link><language>en-us</language><item><title>Dynamic PR environments for services on EKS</title><link>https://anand-kumar.dev/blog/pr-environments-eks-argocd/</link><guid isPermaLink="true">https://anand-kumar.dev/blog/pr-environments-eks-argocd/</guid><description>Per-PR Kubernetes environments on a shared EKS cluster — ArgoCD git-directory generators, a 3-layer Helm value cascade, multi-source Applications, and a namespace-reaper controller for the cleanup ArgoCD pruning misses.</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate><category>gitops</category><category>eks</category><category>argocd</category><category>kubernetes</category><category>helm</category></item><item><title>vLLM vs SGLang vs TGI — operational benchmarks (coming soon)</title><link>https://anand-kumar.dev/blog/vllm-vs-sglang-vs-tgi/</link><guid isPermaLink="true">https://anand-kumar.dev/blog/vllm-vs-sglang-vs-tgi/</guid><description>Side-by-side production benchmarks of the three open-source inference servers — throughput, latency, KV-cache behavior, and the operational gotchas you&apos;ll hit at scale.</description><pubDate>Wed, 20 May 2026 00:00:00 GMT</pubDate><category>inference</category><category>vllm</category><category>sglang</category><category>tgi</category></item></channel></rss>