FixMyCity

A report is only useful if somebody can own the next step.

FixMyCity began with a full operational loop: a citizen submits a photo and location; a municipality reviews the evidence; a crew is assigned; field work returns an after-photo and note; the citizen sees the status change.

Building the whole loop exposed the hard parts that a single reporting form hides: geographic scope, role boundaries, evidence storage, assignment logic, notifications, audit history, and failure states across multiple clients.

One domain, four distinct surfaces.

The Expo mobile app is optimized for reporting and follow-up. The Next.js administration surface is scoped to the active municipality. A NestJS API owns the transactional domain, while a separate FastAPI service isolates image and language analysis. PostgreSQL/PostGIS, Redis, MinIO, and OSRM support spatial queries, queues, media, and routing.

The separation is intentional. AI output can assist classification, severity estimation, duplicate discovery, privacy checks, and text review without becoming the source of truth for municipal action.

Explain the recommendation before asking for trust.

Crew suggestions show the factors behind the score: category expertise, availability, workload, travel distance, and urgency. Routing can fall back when OSRM is unavailable. The interface keeps these limits visible so the operator can make the assignment rather than accept a black box.

The project is a graduation prototype, not a claim of real municipal adoption. Its value is the depth of the end-to-end system and the decisions made explicit inside it.