Our product · Property
Spitometro
Measure before you move.

At a glance
An AI analyst that assembles an evidence-cited dossier on a specific home before you buy, sell or rent.
- An agentic analyst that builds a cited property dossier
- 37 deterministic modules: the model never writes a number
- Payments and myDATA invoicing built in
- Client
- Our own product
- Where
- Greece
- When
- 2026
- Led by
- Fotios Ntovas
- Analysis modules
- 37, all deterministic
- Public datasets
- 40+, each with a freshness contract
- Languages
- Greek · English
- Status
- Pre-launch
01
The brief
What had to be true, and what stood in the way.
Buying a home in Greece usually starts with scattered desk research: planning rules, costs, risks, prices, rents. Spitometro does that groundwork for a specific property in Attica or Thessaloniki. You describe the home in a conversation, in Greek or English, and get a dossier with an indicative value range, legality and planning checks, purchase and holding costs, rental yield, climate and seismic risk, and the neighbourhood, with a source for every figure.
It is built to come before the engineer, valuer or lawyer, not to replace them.
02
Decisions that shaped it
The calls that made the difference, in short. Tap any to read the reasoning.
- 01
An analyst, not a chatbot
Bounded chat sessions share one layered knowledge base per property. The agent explains what it's doing, asks for missing facts with structured prompts, and starts an analysis only when enough modules have what they need. Progress streams live, and analyses re-run as the facts change.
- 02
The model never writes a number
Thirty-seven deterministic modules compute every figure. A claim grammar and copy linters keep the model's prose tied to those numbers and out of regulated advice language. When a source fails, the gap is disclosed, not filled.
- 03
Ranges, honestly calibrated
The indicative value range comes from a gradient-boosting model with calibrated quantile bands. It never sees the asking price, and it is validated with spatially blocked, out-of-time tests and promotion gates.
- 04
Sources that degrade gracefully
Every public dataset has rate limits, circuit breakers, a fallback ladder and a freshness contract, so a slow registry never stalls a report.
- 05
Greek invoicing built in
Payments, receipts and invoices sent to myDATA through a licensed provider, automatic credit notes, VAT-number validation with registry and VIES look-up, the statutory withdrawal flow, and full GDPR export and erasure.
- 06
One report, two forms
The report in the app and the PDF are the same document, printed to paginated A4 by a serverless browser service, in Greek and English.
03
In pictures
Swipe to see more →
04
What we built
Everything we delivered, by discipline.
- Product & brand
- Research, proposition, product design, the house-and-ruler identity.
- Web
- Next.js app with live analysis views and a bilingual report.
- Backend
- Python services, PostgreSQL with PostGIS, 40+ ingestion pipelines.
- AI
- Agent harness on Google ADK and Gemini; the value-range model.
- Commerce
- Stripe payments, myDATA invoicing through Elorus.
Stack
- Next.js
- TypeScript
- Python
- FastAPI
- PostgreSQL
- PostGIS
- LightGBM
- Google ADK
- Gemini
- Cloud Run
- Stripe
- Elorus
Spitometro provides indicative information. It is not a valuation or professional advice.
Ask about this
Questions about Spitometro?
AI answer from Neumatics' assistant. It only knows this site, can make mistakes, and nothing is stored.
Next step
Have something to build?
Tell us about it in a few short steps, then pick a time for a 30-minute call. An idea, a problem or an old site is enough to start.



