Launched this week
Azmoth
Deterministic GOÄ invoice audit, no LLM in the verdict
2 followers
Deterministic GOÄ invoice audit, no LLM in the verdict
2 followers
Azmoth audits German private medical invoices (PADnext, GOÄ) with formal logic, not a language model. Every finding cites the paragraph and the GOÄ wording. Same input, same result, with a hash receipt. Free pilot, synthetic data only.
Free
Launch Team
CloudflareStill paying to deploy your apps? Try Cloudflare Workers for free
Promoted
Hi Product Hunt, I'm Oussama, the maker.
German private doctors bill under the GOÄ fee schedule, and checking an invoice means reading exclusion rules, factor caps and legal notes position by position. Most "AI" tools answer that with a model's best guess. We took the opposite route: formal logic, so the same invoice always gives the same result, and every finding comes with the paragraph and the GOÄ wording next to it.
Every position lands in one of four states: certified wrong, certified correct, a hint your team decides, or "no statement". There is no "looks okay" state, because not finding an error is not proof of correctness. Every report also carries a SHA-256 receipt over catalog, rules, logic, solver version and input, so "deterministic" is something you can recompute.
What it is not, yet: it runs on synthetic test data only. 517 of 2,841 GOÄ items have an enforced rule today (about 18%, as of 2026-10-14); the rest get a hint or "no statement". There is no signed data-processing agreement, no ISO or SOC certification and no pen test. We say that on our own site.
We are looking for pilot partners: billing centres and practice-software vendors in Germany. The pilot is free and non-binding. Questions about the architecture are very welcome.