Slánú Lab (Beta)

Outpatient clinics — especially bilingual and Spanish-speaking practices — need help turning free-text medical notes into billable procedure codes without buying into a big EHR or a proprietary CPT dataset they can't audit. The suggestion has to be explainable, human-reviewable, and priced against the clinic's own fee book.

I designed and built the whole stack solo as a single Next.js 16 app — UI, route handlers, and Prisma access in one process — with all server logic isolated under `lib/server/*` so it can be lifted into a Fastify service later without rewriting it. Cognito + Amplify + aws-jwt-verify gate every API route; `clinicId` is the multi-tenant security boundary, checked in every query. Fee-book upload is alias-aware across English and Spanish column names, and rows upsert on `(clinicId, code)` so re-uploads refresh prices instead of duplicating. The suggestion engine is deterministic — same note in, same codes out — designed so an LLM step can layer on top of it without ever replacing it.