Parashift and onprem.ai, an AI infrastructure company based in Schaffhausen, Switzerland, are now working together. Companies that want to process documents with AI without handing them to a cloud provider can now get the document model Dokumentli on onprem.ai’s on-premise infrastructure. A dedicated GPU project is no longer required.<\/p>\n\n\n\n
In banks, insurers, hospitals, law firms and public administrations, documents frequently contain exactly the data that is most strictly protected: account details, health data, contract terms and personal data. Regulatory requirements, internal policies or customer contracts therefore often rule out processing in a third-party cloud, and even where it would be permitted, many organisations decide against it.<\/p>\n\n\n\n
Until now, running document AI in your own data center meant a separate infrastructure project. Servers with suitable GPUs had to be procured, models installed and kept up to date, and operations and monitoring set up. For many companies, this was the reason to postpone automating their document processes.<\/p>\n\n\n\n
onprem.ai provides servers and software for running AI in-house. Customers own the hardware, and usage is billed at a flat rate that does not depend on the number of tokens processed. Through the partnership, Dokumentli<\/a> runs on this infrastructure and can be integrated directly into existing workflows.<\/p>\n\n\n\n On the server at the customer’s site, Dokumentli recognises the document type, extracts the content and passes it on to downstream systems as structured JSON. Companies that also need review, validation and handover to business applications can connect the Parashift Platform. It runs in the Parashift Cloud with data sovereignty in Switzerland or Germany, receives the results from Dokumentli as JSON in this setup and applies its rule set to them.<\/p>\n\n\n\n \u201cMany of our customers in banking, insurance and healthcare want to process documents with AI, but not hand a single one of them to a third-party cloud. With Dokumentli, we have built a model that runs on a single GPU and outperforms the large language models on documents. With onprem.ai, this performance is now available in your own data center, without requiring an infrastructure project.\u201d<\/p>\nAlain Veuve, CEO and Founder, Parashift<\/cite><\/blockquote>\n\n\n\n \u201cOur customers want to use AI as easily as in the cloud, but on their own hardware. Documents are one of the most common use cases. With Dokumentli, a specialised model comes to our platform that reads invoices, contracts and forms without a single document leaving the building.\u201d<\/p>\nTino B\u00e4chtold, CEO and Founder, onprem.ai<\/cite><\/blockquote>\n\n\n\n In the video (in German), Alain Veuve and Tino B\u00e4chtold talk about the partnership and about how companies process documents with AI on their own hardware.<\/p>\n\n\n\nStatements on the partnership<\/h2>\n\n\n\n
\n
\n
In conversation<\/h2>\n\n\n\n