About Petrarch
Petrarch is a data company that prepares the proprietary operational records of established businesses so that AI developers can use them as training data. Its stated focus is the information held inside economically vital industries, which it argues open datasets and public repositories can no longer match for training frontier models or building long horizon benchmarks.
The work covers the full pipeline. Petrarch sources records from operating businesses, cleans and restructures them, and anonymizes them on the customer's premises for privacy and regulatory compliance. It then extracts what is valuable to build authentic tasks and real world environments that AI systems can learn from. The site frames this as long term partnerships with companies of all sizes rather than one off sales, and says legal compliance and privacy are prioritized at every step.
The navigation lists offerings for AI labs, for special situations and a standalone de identification product, along with a blog, a trust center and a security page. The audience is two sided: businesses that hold valuable records and want them prepared safely, and research labs that need realistic data for training and evaluation. The site invites visitors to request access or book a call.
Petrarch does not publish prices. Engagement begins with a request for access or a call with the founders, and terms are set per partnership.