<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Annotation Radar — H2L Robotics India</title>
    <link>https://h2lrobotics.in/radar/</link>
    <atom:link href="https://h2lrobotics.in/radar/feed.xml" rel="self" type="application/rss+xml"/>
    <description>Curated news on the data annotation and AI training-data industry, with analysis from H2L Robotics India.</description>
    <language>en</language>
    <lastBuildDate>Mon, 16 Nov 2026 09:00:00 GMT</lastBuildDate>
    <item>
      <title>Sama, the AI data-annotation firm, is winding down and liquidating</title>
      <link>https://news.google.com/rss/articles/CBMif0FVX3lxTE9Ed2ppQ2tULUM3WjB5MU05c3FPcHJIdnl0aFNsbkQ1eElCM3NDbHFYRVY1dW5Xbkluc3N4ZHRhYnhjOTdncUFac0dEUE1rSE5NSEp6V1ZwNmh4cmxaRE9KelNFYkoxUU9IdkFreFBQaG9QbHNQRi12SE41X0NXQ00?oc=5</link>
      <guid isPermaLink="false">sama-liquidates-2026-11</guid>
      <pubDate>Mon, 09 Nov 2026 09:00:00 GMT</pubDate>
      <category>Market</category>
      <description>A letter to creditors blames falling revenue from competition, made worse by class-action litigation. For anyone buying annotation, the lesson is supplier risk: how a vendor treats and pays its annotators is not a side issue — it can end the vendor. Stable, salaried in-house teams are a real advantage here.</description>
    </item>
    <item>
      <title>Nvidia in talks to back Mercor at a $20bn valuation</title>
      <link>https://www.marketscreener.com/news/nvidia-mulls-mercor-investment-at-20-billion-valuation-ce7859d2dc88f624</link>
      <guid isPermaLink="false">nvidia-mercor-20b-2026-08</guid>
      <pubDate>Wed, 19 Aug 2026 09:00:00 GMT</pubDate>
      <category>Funding</category>
      <description>The July talks now have a name attached, and it changes the story. If it closes, Mercor passes Scale AI to become the most valuable company in AI data — on a business that barely existed three years ago. Two things sit under the number. Nvidia is reportedly already a customer, paying Mercor to source human-expert data for its Nemotron models, so this is a buyer investing in its own supplier. And the reported $614m first-half revenue leans heavily on a handful of frontier labs — a concentration most annotation businesses would not survive. The money is chasing expert judgement, not labelling volume.</description>
    </item>
    <item>
      <title>MeitY examines home-services data collection for physical-AI training</title>
      <link>https://inc42.com/buzz/how-prontos-physical-ai-pilot-raised-questions-around-consent-surveillance/</link>
      <guid isPermaLink="false">meity-physical-ai-data-2026-11</guid>
      <pubDate>Thu, 12 Nov 2026 09:00:00 GMT</pubDate>
      <category>Policy</category>
      <description>India’s IT ministry is looking at in-home recording pilots used to train robots. Consent and data handling will decide who is allowed to collect real-world data in India — the same questions any annotation workflow on sensitive images has to answer before the work starts.</description>
    </item>
    <item>
      <title>iMerit’s Radha Basu on how AI is reshaping India’s data-annotation jobs</title>
      <link>https://news.google.com/rss/articles/CBMirAFBVV95cUxNb0JtSmU5TkJ1M3RsSzA1OUVVNGgzeFhRUi1OR3VQa1NmdkdlNWNfZUtzLVlLOXJISVJRTVB1Z3NfZVBtNUJJc09RcElxdHhvVFlmcURmTS1aeF9CUjJLWWl0S0h1eEhrcHphRmR3Zkw4QkotREhIWF9VaUNBeXlJZ0pwM2JTYm9tWGdOdno3ZUFObVllMDdRMDlpX3NJR0NqSGdUNlFvS0FHS3Vz?oc=5</link>
      <guid isPermaLink="false">imerit-india-annotation-jobs-2026-11</guid>
      <pubDate>Sat, 07 Nov 2026 09:00:00 GMT</pubDate>
      <category>Market</category>
      <description>The direction is clear from India’s largest players: simple labelling is being automated, and the value moves to annotators who understand the domain. That is the bet we make with crop-disease work — people who know what a sick plant looks like, not just how to draw a polygon.</description>
    </item>
    <item>
      <title>noyb files GDPR complaints against Scale AI over tracking of contractors</title>
      <link>https://www.theregister.com/a/5305757</link>
      <guid isPermaLink="false">noyb-scale-ai-gdpr-2026-10</guid>
      <pubDate>Tue, 27 Oct 2026 09:00:00 GMT</pubDate>
      <category>Policy</category>
      <description>Five complaints in four EU countries about tracking software and ID checks on Outlier contributors; Scale calls the claims unfounded, and no regulator has ruled. Either way, how annotators are monitored is becoming a compliance question, not just an HR one.</description>
    </item>
    <item>
      <title>iMerit launches Scholars, a platform for managed domain-expert work</title>
      <link>https://siliconangle.com/2026/10/26/imerit-launches-platform-to-connect-ai-developers-with-specialized-domain-experts/</link>
      <guid isPermaLink="false">imerit-scholars-2026-10</guid>
      <pubDate>Mon, 26 Oct 2026 09:00:00 GMT</pubDate>
      <category>Product</category>
      <description>A managed service rather than a marketplace: iMerit runs the projects and the quality checks itself. Expert annotation only works with that kind of oversight — which is why consensus and review sit at the centre of how we work.</description>
    </item>
    <item>
      <title>Mecka raises $60M Series B led by Sequoia for robot training data</title>
      <link>https://techfundingnews.com/mecka-ai-raises-60m-from-sequoia-and-nvidia-to-pay-people-to-teach-robots-how-humans-move/</link>
      <guid isPermaLink="false">mecka-60m-2026-10</guid>
      <pubDate>Wed, 07 Oct 2026 09:00:00 GMT</pubDate>
      <category>Funding</category>
      <description>Physical AI needs its own data: people recording everyday tasks so robots can learn how humans move. Field robots are no different — the camera on a weeding machine is only as good as the labelled images behind it.</description>
    </item>
    <item>
      <title>Snorkel AI triples its valuation to $3.5B as demand for training data booms</title>
      <link>https://techcrunch.com/2026/09/22/snorkel-ai-triples-valuation-to-3-5b-as-demand-for-ai-training-data-booms</link>
      <guid isPermaLink="false">snorkel-3-5b-2026-09</guid>
      <pubDate>Tue, 22 Sep 2026 09:00:00 GMT</pubDate>
      <category>Funding</category>
      <description>A $350M Series E, and a shift from labelling software to delivering finished datasets. The market keeps paying for data that is already checked and ready to train on — the output, not the tool.</description>
    </item>
    <item>
      <title>Amazon closes Mechanical Turk on 30 September after 21 years</title>
      <link>https://techrepublic.com/article/news-amazon-mechanical-turk-shutdown/</link>
      <guid isPermaLink="false">aws-mturk-shutdown-2026-09</guid>
      <pubDate>Wed, 26 Aug 2026 09:00:00 GMT</pubDate>
      <category>Market</category>
      <description>The platform that defined crowdsourced labelling is gone. The work did not disappear — it moved to managed teams and expert networks that can be held to a quality standard. Micro-task crowds were cheap; trained annotators with consensus and review are what models now need.</description>
    </item>
    <item>
      <title>Micro1 reaches a $500m gross run rate, up from $7m in a year</title>
      <link>https://techcrunch.com/2026/08/20/ai-data-startup-micro1-reaches-500m-gross-run-rate-amid-ai-training-boom/</link>
      <guid isPermaLink="false">micro1-500m-run-rate-2026-08</guid>
      <pubDate>Thu, 20 Aug 2026 09:00:00 GMT</pubDate>
      <category>Market</category>
      <description>The number that matters is not the run rate but the margin: Micro1 keeps roughly 60–70% of gross, because most of it is paid straight out to contracted doctors, lawyers and engineers. That is a marketplace, not a studio — it scales by recruiting faster, and it owns no method. It also confirms the market is big enough for several players at once rather than consolidating around one, which is the more useful read for anyone specialising.</description>
    </item>
    <item>
      <title>Ex-staff describe “hostile” conditions for the labellers behind Mercor</title>
      <link>https://www.business-humanrights.org/en/latest-news/usa-data-labeling-startup-mercor-faces-allegations-of-hostile-culture-incl-excessive-hours-and-targets/</link>
      <guid isPermaLink="false">mercor-working-conditions-2026-08</guid>
      <pubDate>Tue, 18 Aug 2026 09:00:00 GMT</pubDate>
      <category>Policy</category>
      <description>Reported alongside the funding talks, and worth reading beside them: former employees describe excessive hours and targets across a contracted workforce of around 30,000, and the company says performance-linked bonuses are industry standard. Both things can be true, and that is the point — a marketplace of contractors and a salaried studio are not the same supply chain, and buyers who care where their training data comes from are starting to ask which one they are buying.</description>
    </item>
    <item>
      <title>EXL closes the iMerit acquisition</title>
      <link>https://www.exlservice.com/about/newsroom/exl-completes-acquisition-of-imerit-accelerating-enterprise-ai-leadership</link>
      <guid isPermaLink="false">exl-imerit-closed-2026-08</guid>
      <pubDate>Mon, 03 Aug 2026 09:00:00 GMT</pubDate>
      <category>Acquisition</category>
      <description>The deal we flagged in June has completed, and founder Radha Basu joins EXL&#39;s executive committee rather than leaving after an earnout — which suggests the buyer wants the operating knowledge, not just the delivery capacity. Consolidation in this sector is now a fact rather than a rumour, and the independents left are fewer and more specialised.</description>
    </item>
    <item>
      <title>Scale AI appoints Francis deSouza as chief executive</title>
      <link>https://scale.com/blog/scale-appoints-new-ceo</link>
      <guid isPermaLink="false">scale-ai-new-ceo-2026-07</guid>
      <pubDate>Thu, 30 Jul 2026 09:00:00 GMT</pubDate>
      <category>Market</category>
      <description>An enterprise-software operator from Google Cloud replaces an interim chief, more than a year after founder Alexandr Wang left for Meta. The hire reads as a deliberate turn away from the founder-led, lab-facing era toward selling to ordinary enterprises — which is a different business with different buyers, and it opens room underneath for vendors who still do the specialist work.</description>
    </item>
    <item>
      <title>Mercor in talks to raise $500M at a $20 billion valuation</title>
      <link>https://www.forbes.com/sites/richardnieva/2026/07/09/mercor-fundraise/</link>
      <guid isPermaLink="false">mercor-20b-raise-2026-07</guid>
      <pubDate>Thu, 09 Jul 2026 09:00:00 GMT</pubDate>
      <category>Funding</category>
      <description>The AI-training-data race keeps accelerating — Mercor would double its valuation in under a year, after crossing a reported $2B annualized revenue in June. The money is chasing human expertise in the training loop, not just bulk labeling. Precision, domain-specific annotation is where the margin is moving.</description>
    </item>
    <item>
      <title>EXL to acquire iMerit for up to $310 million</title>
      <link>https://www.globenewswire.com/news-release/2026/06/24/3316801/9060/en/exl-to-acquire-imerit-advancing-its-leadership-in-enterprise-ai-by-adding-foundation-model-expertise-and-technology.html</link>
      <guid isPermaLink="false">exl-acquires-imerit-2026-06</guid>
      <pubDate>Wed, 24 Jun 2026 09:00:00 GMT</pubDate>
      <category>Acquisition</category>
      <description>The biggest annotation-industry consolidation of the year: EXL pays $170M upfront plus up to $140M in earnouts to fold in iMerit&#39;s Ango platform and its expert &#39;Scholars&#39; network for RLHF and model evaluation. It signals that enterprise-AI buyers now want annotation, domain experts and delivery under one roof — good news for specialist studios who can plug into that stack rather than compete with it.</description>
    </item>
    <item>
      <title>Meta redeploys 6,500 engineers to label AI training data</title>
      <link>https://www.techtimes.com/articles/318586/20260617/meta-conscripts-6500-engineers-data-labelers-revolt-exposes-ai-training-ceiling.htm</link>
      <guid isPermaLink="false">meta-engineers-data-labeling-2026-06</guid>
      <pubDate>Wed, 17 Jun 2026 09:00:00 GMT</pubDate>
      <category>Market</category>
      <description>Even a hyperscaler with unlimited compute hit a wall on data quality — and had to put expensive engineers on labeling. The lesson for the whole industry: models are now bottlenecked on high-quality human-labeled data, not GPUs. That&#39;s a structural tailwind for professional annotation.</description>
    </item>
    <item>
      <title>AI data-labeling market projected to reach $6.5B by 2031</title>
      <link>https://www.herohunt.ai/blog/the-ultimate-ai-data-labeling-industry-overview/</link>
      <guid isPermaLink="false">labeling-market-forecast-2026</guid>
      <pubDate>Mon, 01 Jun 2026 09:00:00 GMT</pubDate>
      <category>Research</category>
      <description>Analysts peg the annotation market at ~$2.3B in 2026, nearly tripling by 2031. The growth is real, but it rewards vendors who move up the value chain into specialized, high-accuracy work — which is exactly where a focused agricultural-CV studio wins.</description>
    </item>
    <item>
      <title>Pentagon awards Meta-backed Scale AI a $500 million contract</title>
      <link>https://www.forbes.com/sites/aliciapark/2026/05/06/pentagon-hands-meta-backed-scale-ai-500-million-contract-5-times-last-years-deal-report-says/</link>
      <guid isPermaLink="false">scale-ai-pentagon-500m-2026-05</guid>
      <pubDate>Wed, 06 May 2026 09:00:00 GMT</pubDate>
      <category>Policy</category>
      <description>A fivefold jump on last year&#39;s $100M defense deal. Government and defense are becoming serious annotation buyers, and both demand exactly what agricultural and industrial CV work demands: auditable quality, traceable provenance, and vetted labelers.</description>
    </item>
    <item>
      <title>Encord raises $60M Series C to build the data layer for physical AI</title>
      <link>https://siliconangle.com/2026/02/26/physical-ai-data-infrastructure-startup-encord-lands-60m-accelerate-intelligent-robot-drone-development/</link>
      <guid isPermaLink="false">encord-series-c-60m-2026-02</guid>
      <pubDate>Thu, 26 Feb 2026 09:00:00 GMT</pubDate>
      <category>Funding</category>
      <description>Led by Wellington at a $550M valuation, with revenue from physical-AI customers (robots, drones, autonomous vehicles) up 10x. This is squarely our lane — annotation for machines that act in the real world, not just chatbots. Validates the agricultural-robotics annotation thesis H2L is built on.</description>
    </item>
  </channel>
</rss>
