This year we have built for our client an AI-powered solution that labels incoming medical documents, names then according to business rules and categorize into different folders. Solution is fully Azure-based and hosted in Azure Switzerland North, processing up to 1000 documents a single business day.

We faces many interesting and some times even anecdotical challenges that are worth taking into account for anyone willing to build somethng similar for their organization.

(Should this talk be selected, I would like to bring also customer to the stage with me as a co-speaker to tell about QA and data anonymizaton).

Track for this case can be AI or Azure: I selected Azure, because we use Azure AI Foundry and fully Serverless Azure achitecture, no any oldschool boring tech. Primary focus however will be not on the AI itself - we are not training any models.