Typly Anonimizator removes personal data from documents on your own computer or server. ICF – Intelligent Invoice Centre reads and checks energy invoices for public offices and companies. Both products are built in Poland on top of an NCBR research project.
Detects and permanently removes names, PESEL numbers, addresses and bank account numbers from PDF, Word and scanned documents. Runs locally: on a computer (Windows, macOS) or on your company's server. It also lets you use ChatGPT and similar tools safely on client documents.
The system reads electricity invoices, including those from Poland's KSeF e-invoicing system, splits them into components and shows the charges hidden behind the total. Running in production at a municipal office.
For a leading real-estate company listed on the Warsaw Stock Exchange we migrated over a thousand documents from a legacy file-share to SharePoint, with full AI classification: document type, counterparty, category, descriptive file name and business metadata — without a single user click.
Project co-financed by the Polish National Centre for Research and Development (NCBR) under the strategic programme „Advanced information, telecommunications and mechatronic technologies" — INFOSTRATEG IV. The aim of the project was to develop an intelligent correspondence dispatcher for the Polish public and business sectors, integrated with EPUAP and EDZ RP.
The R&D outcomes from this project power our commercial products — including ICF – Intelligent Invoice Centre and Typly Anonimizator.
urzad.ai · See the public-administration deployment → Typly Mail · The correspondence dispatcher developed in the project →We advise, design and deploy AI solutions for businesses and public institutions. End-to-end stack: from training our own models, through LLM integrations, to data migrations and process automation.
Windows Share → SharePoint, Confluence, Google Workspace. Classification, descriptive names, metadata — all automatic.
Fine-tuning, RAG, n-gram models, embeddings. For legal, energy, medical and other verticals.
LLM chatbots, voice agents, WhatsApp / Slack / Teams integrations. Support in Polish + 40 languages.
Correspondence classification, data extraction, query dispatcher. Measurable cost reduction.
Deployments inside your own infrastructure, with locally hosted models. Zero data leaving the EU.
Vector store, observability, model evaluation, GitOps on k8s. Advisory and turn-key delivery.
Reply suggestions, text correction and translation in 41 languages, with word prediction running offline.