<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Backend on Top AI Skills</title><link>https://topaiskills.com/tags/backend/</link><description>Recent content in Backend on Top AI Skills</description><generator>Hugo</generator><language>en</language><lastBuildDate>Fri, 24 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://topaiskills.com/tags/backend/index.xml" rel="self" type="application/rss+xml"/><item><title>Groq API: Ultra-Fast LLM Inference for AI Agents</title><link>https://topaiskills.com/skills/coding/groq-api/</link><pubDate>Fri, 24 Jul 2026 00:00:00 +0000</pubDate><guid>https://topaiskills.com/skills/coding/groq-api/</guid><description>&lt;h2 id="what-it-is"&gt;What It Is&lt;/h2&gt;
&lt;p&gt;Groq is not another cloud GPU provider — it is a hardware company that built a custom Language Processing Unit (LPU) from the ground up. Where traditional GPUs like NVIDIA&amp;rsquo;s H100 or AMD&amp;rsquo;s MI300X process tokens in batches with millisecond-level latency, the Groq LPU is a deterministic architecture designed for one job: running transformer models at memory-bandwidth speeds.&lt;/p&gt;
&lt;p&gt;The result is staggering. Open-source models like Llama 3, Mixtral, and Gemma run at &lt;strong&gt;300–1000+ tokens per second&lt;/strong&gt; on Groq&amp;rsquo;s API — roughly 5–10× faster than the same models on conventional GPU infrastructure. A 2,000-token response that takes 8 seconds on a standard API finishes in under 2 seconds on Groq.&lt;/p&gt;</description></item><item><title>WhatsApp Cloud API</title><link>https://topaiskills.com/skills/coding/whatsapp-cloud-api/</link><pubDate>Thu, 23 Jul 2026 00:00:00 +0000</pubDate><guid>https://topaiskills.com/skills/coding/whatsapp-cloud-api/</guid><description>&lt;h2 id="overview"&gt;Overview&lt;/h2&gt;
&lt;p&gt;The WhatsApp Cloud API skill is a structured reference for AI coding assistants that need to work with Meta&amp;rsquo;s official WhatsApp Business Platform. Instead of digging through Facebook developer docs every time you need to send a message, upload media, or verify a webhook, your agent loads the relevant specification on demand — with real payload examples, error codes, and best practices baked in.&lt;/p&gt;
&lt;p&gt;This skill follows a &lt;strong&gt;progressive disclosure&lt;/strong&gt; design. It loads only ~100 tokens at startup (just the skill name and a brief description). When your agent detects WhatsApp-related work, it loads the main reference (~4500 tokens). Specific topic files — messaging, templates, webhooks, media — stay at zero tokens until your agent actually needs them. The result is a comprehensive reference that doesn&amp;rsquo;t bloat your context window.&lt;/p&gt;</description></item></channel></rss>