<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0">
    <channel>
      <title>Egor Dmitriev</title>
      <link>https://egordmitriev.dev</link>
      <description>Last 10 notes on Egor Dmitriev</description>
      <generator>Quartz -- quartz.jzhao.xyz</generator>
      <item>
    <title>Inside Gradient Boosting</title>
    <link>https://egordmitriev.dev/series/inside-gradient-boosting</link>
    <guid>https://egordmitriev.dev/series/inside-gradient-boosting</guid>
    <description><![CDATA[ A deep dive into how XGBoost and LightGBM work - from first principles to implementation details. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Optimizing SAMURAI</title>
    <link>https://egordmitriev.dev/series/optimizing-samurai</link>
    <guid>https://egordmitriev.dev/series/optimizing-samurai</guid>
    <description><![CDATA[ Taking a video segmentation model from 1.3 FPS to real-time. ONNX export, TensorRT, CoreML, and Rust. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Research Passes</title>
    <link>https://egordmitriev.dev/series/research-passes</link>
    <guid>https://egordmitriev.dev/series/research-passes</guid>
    <description><![CDATA[ Surveys of a field before deciding to build in it: what exists, what doesn't, and where it's heading. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Blog</title>
    <link>https://egordmitriev.dev/blog/</link>
    <guid>https://egordmitriev.dev/blog/</guid>
    <description><![CDATA[ A collection of articles about machine learning, data engineering, and software development. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Egor Dmitriev</title>
    <link>https://egordmitriev.dev/</link>
    <guid>https://egordmitriev.dev/</guid>
    <description><![CDATA[ Projects Boosters - A high-performance gradient boosting library for Python and Rust Research - Public pre-build surveys, bibliographies, and build/no-build decisions for projects I am considering Series Research Passes - Structured surveys of a research-and-engineering landscape before committing t... ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Boosters</title>
    <link>https://egordmitriev.dev/projects/boosters</link>
    <guid>https://egordmitriev.dev/projects/boosters</guid>
    <description><![CDATA[ A high-performance gradient boosting library for Python and Rust ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Projects</title>
    <link>https://egordmitriev.dev/projects/</link>
    <guid>https://egordmitriev.dev/projects/</guid>
    <description><![CDATA[ Projects A showcase of my open-source work and side projects. Featured 🚀 Boosters A high-performance gradient boosting library for Python and Rust. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Research</title>
    <link>https://egordmitriev.dev/projects/research</link>
    <guid>https://egordmitriev.dev/projects/research</guid>
    <description><![CDATA[ Pre-build surveys and reference notes for projects I am considering building, plus the artifacts that survive when I decide not to build them. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Series</title>
    <link>https://egordmitriev.dev/series/</link>
    <guid>https://egordmitriev.dev/series/</guid>
    <description><![CDATA[ A collection of multi-part blog series exploring technical topics in depth. ]]></description>
    <pubDate>Sun, 12 Jul 2026 20:28:38 GMT</pubDate>
  </item><item>
    <title>Did Your Agent Actually Read That File?</title>
    <link>https://egordmitriev.dev/blog/2026-07-12-did-your-agent-actually-read-that-file</link>
    <guid>https://egordmitriev.dev/blog/2026-07-12-did-your-agent-actually-read-that-file</guid>
    <description><![CDATA[ Your coding agent read twelve files and gave you an answer. Which ones did it actually use? Ablation attribution answers this with ~64 forward passes and a linear regression, and it is more elegant than it has any right to be. ]]></description>
    <pubDate>Sat, 11 Jul 2026 22:00:00 GMT</pubDate>
  </item>
    </channel>
  </rss>