<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><atom:link href="https://datadhara.in/blog/feed.xml" rel="self" type="application/rss+xml"/><atom:link href="https://pubsubhubbub.appspot.com/" rel="hub"/><title>Datadhara blog</title><link>https://datadhara.in/blog</link><description>Real-world data for AI training: licensing, collection and de-identification.</description><language>en</language><item><title>How to license real-world training data for AI</title><link>https://datadhara.in/blog/how-to-license-ai-training-data</link><guid>https://datadhara.in/blog/how-to-license-ai-training-data</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>A step-by-step walkthrough for AI teams licensing private, real-world data: writing the spec, running a pilot, reading the dataset card, and negotiating license terms and pricing.</description></item><item><title>De-identifying clinical notes for AI training: what HIPAA Safe Harbor requires</title><link>https://datadhara.in/blog/de-identifying-clinical-notes-hipaa-safe-harbor</link><guid>https://datadhara.in/blog/de-identifying-clinical-notes-hipaa-safe-harbor</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>HIPAA&#x27;s Safe Harbor method lists 18 identifiers that must be removed before health data counts as de-identified. Here is what that means for free-text clinical notes used in AI training, and when Expert Determination fits better.</description></item><item><title>Code-mixed speech data: why ASR breaks when speakers switch languages</title><link>https://datadhara.in/blog/code-mixed-speech-data-for-asr</link><guid>https://datadhara.in/blog/code-mixed-speech-data-for-asr</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>Many speakers switch languages mid-sentence. Here is why speech recognition models struggle with code-mixed audio, what good code-mixed training data looks like, and how to collect it with consent.</description></item><item><title>Handwritten invoices and bills: the training data gap in document AI</title><link>https://datadhara.in/blog/handwritten-invoice-data-for-document-ai</link><guid>https://datadhara.in/blog/handwritten-invoice-data-for-document-ai</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>Document AI models handle clean printed invoices well and struggle with handwritten bills, phone photos and crossed-out lines. Here is why, what labels matter, and how to build a dataset of real ones.</description></item><item><title>Synthetic or real-world training data? Where each one works</title><link>https://datadhara.in/blog/synthetic-vs-real-world-training-data</link><guid>https://datadhara.in/blog/synthetic-vs-real-world-training-data</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>Synthetic data is fast and cheap; real-world data is slow and messy. A practical look at where each works for AI training and evaluation, and why the two work best together.</description></item><item><title>A dataset card template for licensed AI training data</title><link>https://datadhara.in/blog/dataset-card-template</link><guid>https://datadhara.in/blog/dataset-card-template</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>A free, copyable dataset card template for licensed AI training data, covering source, consent, collection, labelling, de-identification, license terms and known gaps.</description></item><item><title>Where to get AI training data that isn&#x27;t on the public web</title><link>https://datadhara.in/blog/where-to-get-ai-training-data-not-on-the-web</link><guid>https://datadhara.in/blog/where-to-get-ai-training-data-not-on-the-web</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>The main routes to non-public AI training data compared: data marketplaces, annotation vendors, custom collection, direct partnerships and your own customers&#x27; data. What each is good for, and the questions to ask.</description></item><item><title>Data provenance for AI training: what it is and how to document it</title><link>https://datadhara.in/blog/data-provenance-for-ai-training</link><guid>https://datadhara.in/blog/data-provenance-for-ai-training</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>Data provenance is the documented origin and history of every record in a training set. What it covers, why buyers and regulators ask for it, and a practical way to record it per record.</description></item><item><title>Retail POS and inventory data for demand forecasting models</title><link>https://datadhara.in/blog/retail-pos-data-for-demand-forecasting</link><guid>https://datadhara.in/blog/retail-pos-data-for-demand-forecasting</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>What makes point-of-sale and inventory data useful for training demand forecasting and retail AI models: granularity, stock-outs, promotions, store diversity and how to license it from independent retailers.</description></item><item><title>How AI training data is priced</title><link>https://datadhara.in/blog/how-ai-training-data-is-priced</link><guid>https://datadhara.in/blog/how-ai-training-data-is-priced</guid><pubDate>Wed, 18 Nov 2026 00:00:00 +0000</pubDate><description>The common pricing models for licensed AI training data (per record, per audio hour, per project, subscriptions) and the factors that move the price: exclusivity, labelling depth, source access and de-identification.</description></item><item><title>Selling a shut-down startup&#x27;s code and data to AI labs</title><link>https://datadhara.in/blog/sell-startup-codebase-and-data-after-shutdown</link><guid>https://datadhara.in/blog/sell-startup-codebase-and-data-after-shutdown</guid><pubDate>Fri, 20 Nov 2026 00:00:00 +0000</pubDate><description>What founders of shut-down startups can license to AI labs (codebases, Slack, tickets, CRM, docs, financials), who controls the IP after a shutdown, what to strip out, and how the process works.</description></item><item><title>How to license your company&#x27;s data to AI companies, step by step</title><link>https://datadhara.in/blog/how-to-license-company-data-to-ai-companies</link><guid>https://datadhara.in/blog/how-to-license-company-data-to-ai-companies</guid><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><description>A step-by-step guide for companies licensing internal data to AI labs: what to inventory, how to confirm you have the rights, what to leave out, what goes in the licence, and how data is cleaned and delivered.</description></item><item><title>What is your startup&#x27;s codebase worth to AI labs?</title><link>https://datadhara.in/blog/what-is-your-codebase-worth-to-ai-labs</link><guid>https://datadhara.in/blog/what-is-your-codebase-worth-to-ai-labs</guid><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><description>What makes a private codebase valuable to AI labs for training and evaluating coding agents: never-public code, real git history, reviews, tests, context around the code, clean ownership and the licence terms.</description></item><item><title>Selling data and IP when a startup shuts down: what founders should know</title><link>https://datadhara.in/blog/selling-startup-data-and-ip-after-shutdown-us-uk-eu</link><guid>https://datadhara.in/blog/selling-startup-data-and-ip-after-shutdown-us-uk-eu</guid><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><description>Who can sell or license a startup&#x27;s code and data during a shutdown, and which approvals apply: board and investor consent, ABCs, bankruptcy, administration and liquidation in the US, UK and EU.</description></item><item><title>Is it legal to sell employee Slack and email data for AI training?</title><link>https://datadhara.in/blog/is-it-legal-to-sell-employee-slack-email-data-for-ai-training</link><guid>https://datadhara.in/blog/is-it-legal-to-sell-employee-slack-email-data-for-ai-training</guid><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><description>What the GDPR, UK GDPR, CCPA and FTC guidance mean for companies licensing workplace Slack, Teams and email to AI labs: lawful basis, employee consent, transparency, anonymisation versus pseudonymisation, and a careful process.</description></item></channel></rss>
