<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
<channel>
  <title>Skevyr Innovations blog</title>
  <link>https://skevyr.com/blog/</link>
  <description>Skevyr Innovations builds the data that AI learns from: data collection, annotation, AI training and model evaluation, delivered by a vetted network of freelancers, domain experts and vendor teams.</description>
  <language>en-in</language>
<item>
  <title>What is RLHF, and why expert human feedback matters</title>
  <link>https://skevyr.com/blog/what-is-rlhf-and-why-expert-feedback-matters/</link>
  <guid>https://skevyr.com/blog/what-is-rlhf-and-why-expert-feedback-matters/</guid>
  <pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
  <category>AI training</category>
  <description>Large language models learn to be helpful from people who judge their answers. Who those people are decides what the model learns.</description>
</item>
<item>
  <title>Building speech datasets for Indian languages: what it really takes</title>
  <link>https://skevyr.com/blog/building-speech-datasets-for-indian-languages/</link>
  <guid>https://skevyr.com/blog/building-speech-datasets-for-indian-languages/</guid>
  <pubDate>Thu, 24 Sep 2026 00:00:00 GMT</pubDate>
  <category>Data collection</category>
  <description>Hours of audio are easy to count. Useful hours, with the right speakers, conditions and consent, take careful design.</description>
</item>
<item>
  <title>Five layers of annotation quality: how to make labels you can trust</title>
  <link>https://skevyr.com/blog/five-layers-of-annotation-quality/</link>
  <guid>https://skevyr.com/blog/five-layers-of-annotation-quality/</guid>
  <pubDate>Tue, 15 Sep 2026 00:00:00 GMT</pubDate>
  <category>Annotation</category>
  <description>Annotation quality isn't a final inspection. It's a system of checks running from the first label to the last delivery.</description>
</item>
</channel>
</rss>