Five layers that work as one: workbenches where linguists build the lexicon, a pipeline that compiles it, an engine that serves it, the apps that deliver it — and the platform that runs the whole programme.
Two workbenches hold everything a language knows about itself. Linguists build each entry once; every product reads from it.
Where lexicographers create and maintain the lexicon.
Where the grammar of every word is modelled and validated.
Deep enough for a richly inflected language.
These are the production figures of the Albanian deployment, counted when the data was migrated to the new platform on 1 October 2026.
Including homonyms, with part of speech and gender
Definitions written and reviewed by lexicographers
Plus 6,317 antonym links, per meaning
Real sentences showing each meaning in use
Plus 5,180 proverbs
Paradigm cells for nouns, adjectives and verbs
Domain labels from law to medicine
Across 46 tables, migrated in one cutover
The export pipeline turns the factory's work into production data in phases: headwords and meanings; examples, proverbs and phraseology; translations and links; synonyms and antonyms; and finally word forms and inflection tables.
Small, isolated services with one job each — so they scale independently and fail safely.
Prefix search from an in-memory index that forgives missing diacritics; entries, homonyms and inflection tables; a fast “is this a word?” check for extensions; rate-limited public word pages for search engines; word-of-the-day scheduling.
Spelling, grammar, style and synonyms behind a LanguageTool-compatible interface. Runs privately with no internet exposure; results are cached so unchanged paragraphs are never re-checked; updates are smoke-tested on a real sentence and rolled back on failure.
Rewrites a sentence in professional, formal or friendly tone using a large language model — and every rewrite is re-checked by the proofing engine before it reaches the writer.
Neural voices read headwords, single meanings and any text. Clips are cached per voice, each plan has its own voices and allowance, and administrators can tune pronunciation instructions.
Cloud documents with folders, trash and storage quotas, sharing and comments, Word import and export, PDF export — and real-time co-editing for teams.
04 · Deliver — fifteen products on eleven platforms, all reading from this engine.
See the productsThe services that turn a set of apps into a sustainable programme — identity, licensing, communications, growth, content and administration — shared by every app.
One account for every app. Email, Google, Apple and Facebook sign-in, authenticator-app two-factor, device limits — and account deletion that erases a person's data across every service.
Free, government-funded, professional and premium plans. Team seats, gift subscriptions, referrals, discount codes and admin grants, with hosted checkout and tax handling.
Event-driven automations, templates and campaigns across email, push notifications and an in-app inbox — with unsubscribes and suppression lists handled for you.
A B2B pipeline for schools, ministries and companies, ad and analytics connectors, user cohorts and alerts — and AI-assisted analysis that waits for human approval.
Every interface string of every app and every page of the website, managed centrally, versioned, with full history.
Users, plans, payments, messaging, speech, dictionary and content — run from one dashboard, with roles for operators and marketing.
Built on widely used, well-supported technology, so your own engineers can understand, operate and audit it.
Most of the ecosystem is language-independent and ready on day one. Some of it is configured for your language with your linguists. And some of it only your linguists can build. We show you exactly which is which before you commit.
Language-independent — works as is
Adapted by our engineers with your linguists
Your language's knowledge, in your institution's hands
A national language is national infrastructure. The ecosystem is engineered to be hosted where you decide, private by default and accountable in everything it does.
Every service is containerised and orchestrated with Kubernetes, so the whole ecosystem can run in your own national data centre — or be operated by us.
Internal services and the proofing engine are never exposed to the internet, and the engine has no outbound network access at all.
Deleting an account triggers erasure in every service. Automated tests block any new personal-data field that erasure does not cover.
AI rewrites are re-checked by the proofing engine. AI-proposed growth actions sit in an approval queue, with an emergency stop.
Separate staging and production environments, gated releases, and proofing-engine updates that test a real sentence and roll back on failure.
Metrics dashboards and centralised logs across services, so operators see problems before citizens do.
We walk your technical team through the platform — the workbenches, the pipeline, the engine and the operations behind them.