Product Database with Automated Claim Checking
Product information lived across the shop, a grown series of Excel versions, CSV exports and Notion. At the same time, health claims on food are strictly regulated — one wrong sentence is a legal risk.
Projects
Applications in production, not prototypes in a drawer.
Each of these solved a real problem in the day-to-day work of a company or in my own. Where a system writes outwards, a verifiable rule sits in front of the model — and everything here runs on infrastructure I set up and maintain myself.
Marketing Dashboard
Four separate tools, and the simplest question — is our marketing actually producing revenue? — could not be answered without half a day of clicking exports together.
Complaint Assistant
Complaints arrived as free text and were transferred by hand. Now a pipeline collects them, has a language model classify them, and creates ready-made tickets — with the assessment attached as an internal note.
Chatty — a workspace for long AI conversations
In long conversations with language models the beginning gets lost, and good material from old chats is impossible to find again. So I built myself an interface.
Influencer Matcher
Over a hundred collaboration partners each needed the right product assigned. The basis for it — years of collaboration history — existed, but only as a mess of spreadsheets.
Voice Assistant
Talking instead of typing. Later, one component was carved out as its own secured endpoint so another project could use just the part it needed.