<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>mattis megevand</title>
    <link>https://mattismegevand.com/</link>
    <description>mattis personal website</description>
    <atom:link href="https://mattismegevand.com/rss.xml" rel="self" type="application/rss+xml" />
    <lastBuildDate>Tue, 24 Feb 2026 21:59:55 GMT</lastBuildDate>
    <item>
      <title>repere - exploring data without leaving the browser</title>
      <link>https://mattismegevand.com/posts/repere/</link>
      <guid>https://mattismegevand.com/posts/repere/</guid>
      <description>a local-first data explorer with shareable, replayable pipelines that runs sql entirely client-side using duckdb-wasm</description>
    </item>
    <item>
      <title>a unified method for neural network compression</title>
      <link>https://mattismegevand.com/posts/neural-network-compression/</link>
      <guid>https://mattismegevand.com/posts/neural-network-compression/</guid>
      <description>making deep learning practical on embedded hardware by letting networks learn their own compression quantization, pruning, and low-rank factorization in one training loop</description>
    </item>
    <item>
      <title>CMS-like editing for Markdown with contenteditable and 100 lines of JS</title>
      <link>https://mattismegevand.com/posts/inlinecms/</link>
      <guid>https://mattismegevand.com/posts/inlinecms/</guid>
      <description>A minimal dev-only CMS for inline editing of Markdown content with live preview and automatic file saving</description>
    </item>
    <item>
      <title>invertornot - predicting image invertibility for dark mode</title>
      <link>https://mattismegevand.com/posts/invertornot/</link>
      <guid>https://mattismegevand.com/posts/invertornot/</guid>
      <description>a small neural network that decides whether images should be inverted in dark mode, trained on real user feedback and deployed as a fast api</description>
    </item>
    <item>
      <title>sqlgbm - running ml models directly in sql</title>
      <link>https://mattismegevand.com/posts/sqlgbm/</link>
      <guid>https://mattismegevand.com/posts/sqlgbm/</guid>
      <description>converting gradient boosting models to sql queries for zero latency inference without additional infrastructure</description>
    </item>
  </channel>
</rss>
