Autofract

Case — inside the system

Moving countries, minus the guesswork.

Open data turned into calculators, corridors and a live community — in 22 languages.

50

countries in the dataset core

33

calculators, all client-side

22

languages, one pipeline

0

runtime servers to break

Relocating starts where the anxiety starts: what will it actually cost, what visa fits, how long until citizenship. A build-time ETL normalizes World Bank, weather and geographic data into 30+ datasets across 50 countries; a quality gate fails the entire build if a single dataset is inconsistent. From that spine, a page engine generates country corridors, comparisons and rankings, and 33 calculators run entirely client-side.

The whole site is then mirrored into 22 languages by its own translation pipeline — self-hosted models, a versioned translation memory, and a hard build failure on any untranslated string. It ships as a fully static build to the edge: nothing to crash at runtime. Alongside the data lives City Chat, where every message passes AI moderation before it goes public, backed by EU-resident Postgres with row-level security and GDPR retention by construction.

System map

Live simulation — hover the nodes, drag them around.

Core AI Workers Data Delivery Gates Surface

Inside the system

Relocating

Open data turned into a relocation product

Open-Data ETL

World Bank, weather, geo — normalized per country

Dataset Core

30+ datasets covering 50 countries

Quality Gate

Bad data fails the build — literally

Page Engine

Programmatic corridor, compare and ranking pages

Tool Suite

33 calculators, all running client-side

Static Build

Fully prebuilt — no runtime servers to break

i18n Pipeline

Mirrors the whole site into 22 languages

Translation Memory

Versioned cache — deploys never call live MT

MT Engine

Self-hosted translation models, GPU passes

AI Discovery

llms.txt, structured data — built to be cited

Edge Delivery

Static edge, auto-TLS, immutable caching

City Chat

Realtime rooms with presence, seeded by data

AI Moderation

Every message reviewed by a model before it’s public

Trust & Safety

Bans, audit log, rate limits, GDPR retention

EU Postgres

Row-level security everywhere, minimal PII

Want a system like this?

This case was designed, built and operated by the studio, end to end. Yours can be next — fixed quote, production from week one.

Hire the studio

Next system

PathCore →