<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Data For Science: LLM engineering articles</title><description>Technical articles on LLMs, RAG, agents, evaluation, and self-hosting from Data For Science, each with companion code on GitHub.</description><link>https://data4sci.com/blog</link><language>en</language><atom:link href="https://data4sci.com/blog/rss.xml" rel="self" type="application/rss+xml"/><item><title>Jev 101</title><link>https://data4sci.com/blog/jev-101</link><guid isPermaLink="true">https://data4sci.com/blog/jev-101</guid><description>A new class of model. In this post, we take a look at Jev, a new “System 1” type of model optimized for extremely fast and cheap decision making with…</description><pubDate>Tue, 03 Nov 2026 00:00:00 GMT</pubDate><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>Local LLM Inference at Scale with vLLM</title><link>https://data4sci.com/blog/local-llm-inference-at-scale-with</link><guid isPermaLink="true">https://data4sci.com/blog/local-llm-inference-at-scale-with</guid><description>Continuous batching, prefix caching, and structured outputs: the vLLM features that turn a single local model into a serving engine for thousands of requests.</description><pubDate>Fri, 09 Oct 2026 00:00:00 GMT</pubDate><category>LLM serving</category><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>The Epidemiology of a Cognitive Virus</title><link>https://data4sci.com/blog/the-epidemiology-of-a-cognitive-virus</link><guid isPermaLink="true">https://data4sci.com/blog/the-epidemiology-of-a-cognitive-virus</guid><description>What happens when you take &quot;LLM addiction&quot; seriously enough to write down the equations and run simulations. In the latest epidemiology post, we are…</description><pubDate>Mon, 21 Sep 2026 00:00:00 GMT</pubDate><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>LLMs Are the New Wikipedia</title><link>https://data4sci.com/blog/llms-are-the-new-wikipedia</link><guid isPermaLink="true">https://data4sci.com/blog/llms-are-the-new-wikipedia</guid><description>Every old criticism is new again. Welcome to the latest post on the LLM series. In this post, we take a different tack and take a broader view of the LLM…</description><pubDate>Sun, 06 Sep 2026 00:00:00 GMT</pubDate><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>An MCP Server from Scratch</title><link>https://data4sci.com/blog/an-mcp-server-from-scratch</link><guid isPermaLink="true">https://data4sci.com/blog/an-mcp-server-from-scratch</guid><description>Interact with a database with the latest version of the MCP Protocol. In this post we build a custom MCP server from scratch using nothing but the Python…</description><pubDate>Tue, 25 Aug 2026 00:00:00 GMT</pubDate><category>AI agents</category><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>Evaluating your Agentic Harnesses</title><link>https://data4sci.com/blog/evaluating-your-agentic-harnesses</link><guid isPermaLink="true">https://data4sci.com/blog/evaluating-your-agentic-harnesses</guid><description>One good demo is a test flight. An eval suite is the flight-test campaign: pass rates, cost, latency, and failure modes.</description><pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate><category>LLM evaluation</category><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>Self hosting LLMs with Llama.cpp</title><link>https://data4sci.com/blog/self-hosting-llms-with-llamacpp</link><guid isPermaLink="true">https://data4sci.com/blog/self-hosting-llms-with-llamacpp</guid><description>Install, download, and serve a small model on Apple Silicon, then read the model file&apos;s guts — including the arithmetic for working out whether a given…</description><pubDate>Tue, 28 Jul 2026 00:00:00 GMT</pubDate><category>LLM serving</category><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>Building an Advanced Agentic Harness</title><link>https://data4sci.com/blog/building-an-advanced-agentic-harness</link><guid isPermaLink="true">https://data4sci.com/blog/building-an-advanced-agentic-harness</guid><description>From a single pilot to an air campaign: planning, parallelism, memory, verification, and observability for production-shaped agents.</description><pubDate>Wed, 15 Jul 2026 00:00:00 GMT</pubDate><category>AI agents</category><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>Build a (small) language model by counting</title><link>https://data4sci.com/blog/build-a-small-language-model-by-counting</link><guid isPermaLink="true">https://data4sci.com/blog/build-a-small-language-model-by-counting</guid><description>A 4-gram model on WikiText-103, and the one idea it shares with Claude. This week we continue our exploration of Large Language Models by building a simple…</description><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><author>info@data4sci.com (Bruno Gonçalves)</author></item><item><title>Building a Basic Agentic Harness</title><link>https://data4sci.com/blog/building-a-basic-agentic-harness</link><guid isPermaLink="true">https://data4sci.com/blog/building-a-basic-agentic-harness</guid><description>The secret sauce behind Claude Code, Open Clawd, Hermes, etc. Every coding agent you’ve ever used: Claude Code, OpenClawd, Cursor’s agent mode, Devin…</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate><category>AI agents</category><author>info@data4sci.com (Bruno Gonçalves)</author></item></channel></rss>