<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Abhijeet Solanki · Blog</title>
    <link>https://abhijeetsolanki.com/blog</link>
    <atom:link href="https://abhijeetsolanki.com/rss.xml" rel="self" type="application/rss+xml"/>
    <description>Notes on machine learning, AI security and building things.</description>
    <language>en</language>
    <lastBuildDate>Mon, 28 Sep 2026 12:00:00 GMT</lastBuildDate>
    <item>
      <title>Building Prooftrace: an AI tester that has to prove itself</title>
      <link>https://abhijeetsolanki.com/blog/building-prooftrace</link>
      <guid isPermaLink="true">https://abhijeetsolanki.com/blog/building-prooftrace</guid>
      <pubDate>Mon, 28 Sep 2026 12:00:00 GMT</pubDate>
      <description>Acceptance criteria are still checked by hand, release after release. Here is how I built an AI tester that drives a real browser, proves each criterion, and hands back tests your team keeps, and the rule that keeps it honest: no proof, no pass.</description>
      <category>AI agents</category>
      <category>Testing</category>
      <category>Playwright</category>
    </item>
    <item>
      <title>Building Voice Auditor: the polite call that broke two rules</title>
      <link>https://abhijeetsolanki.com/blog/building-voice-auditor</link>
      <guid isPermaLink="true">https://abhijeetsolanki.com/blog/building-voice-auditor</guid>
      <pubDate>Mon, 28 Sep 2026 12:00:00 GMT</pubDate>
      <description>A support call scored 99.9 out of 100 and still broke two company rules. Here is how I built Voice Auditor to check every call against written policy, what the first version got wrong, and how it keeps card numbers out of the database.</description>
      <category>Speech AI</category>
      <category>Compliance</category>
      <category>Privacy</category>
    </item>
    <item>
      <title>Detecting the Odd One Out: A Guide to One-Class Classification</title>
      <link>https://abhijeetsolanki.com/blog/detecting-the-odd-one-out</link>
      <guid isPermaLink="true">https://abhijeetsolanki.com/blog/detecting-the-odd-one-out</guid>
      <pubDate>Wed, 08 Jan 2025 12:00:00 GMT</pubDate>
      <description>How one-class classifiers learn what normal looks like, when to use them, and a one-class SVM you can run in a few lines.</description>
      <category>Machine learning</category>
      <category>Anomaly detection</category>
    </item>
    <item>
      <title>How to Start with Data Science in 2020</title>
      <link>https://abhijeetsolanki.com/blog/how-to-start-with-data-science-in-2020</link>
      <guid isPermaLink="true">https://abhijeetsolanki.com/blog/how-to-start-with-data-science-in-2020</guid>
      <pubDate>Sun, 08 Nov 2020 12:00:00 GMT</pubDate>
      <description>What data science is, the skills it takes, and the courses, books and podcasts I'd use to get started.</description>
      <category>Data science</category>
      <category>Careers</category>
    </item>
    <item>
      <title>What Will Your Plate Look Like in 2050?</title>
      <link>https://abhijeetsolanki.com/blog/what-will-your-plate-look-like-in-2050</link>
      <guid isPermaLink="true">https://abhijeetsolanki.com/blog/what-will-your-plate-look-like-in-2050</guid>
      <pubDate>Mon, 13 Jan 2020 12:00:00 GMT</pubDate>
      <description>Population growth, soil degradation and a financing gap threaten the world's food supply, and what vertical farming, soil sensors and better financing could change.</description>
      <category>Technology</category>
      <category>Agriculture</category>
    </item>
  </channel>
</rss>
