Metatron Production · AI consulting
Not a generic chat tool, but a system that works for your business: it remembers what was agreed, runs on your infrastructure, checks answers before delivery and waits for your confirmation before anything irreversible.
Design, build and long-term operation. In production for institutions and schools. Belgrade.
Transparent AI: the interface always makes it clear when the other party is not a human.
Generic versus personalized
| Criterion | Personalized system (Metatron Production) | Generic cloud assistant |
|---|---|---|
| Whose process | Designed around your workflow: who sends the request, who decides, what is allowed, what is irreversible — with acceptance criteria that are measured. | One pattern for everyone; the work adapts to the tool. |
| Where the data is | On your infrastructure; a local model where data never leaves, cloud only at your choice. | Goes to a third party that processes it. |
| What it remembers | Persistently, in readable files outside the model, with dates; swapping the model does not erase the knowledge. | A session, or an account with a single vendor. |
| Who checks | The model's output is a candidate: it is checked before delivery, and an irreversible action waits for a human to confirm. | The model's answer arrives directly. |
| Voice | Speech transcription and Serbian speech synthesis run on your infrastructure. | The voice recording goes to the vendor. |
| Who chooses the model | You: an explicit choice, no silent substitution; unavailability returns a clear error. | The vendor chooses the version and the rules. |
The comparison is qualitative and refers to local mode. A cloud model is used at your choice, with personal data removed when you require it.
What the house builds
Every service rests on the same core: it runs on any model, checks output before delivery and remembers even when the model is swapped.
Workflow capture, a system blueprint and a work order with acceptance criteria — before anything is built.
One core that remembers: a local model on your hardware or a cloud model at your choice, with output checked before delivery.
Work done on your behalf — reading and writing files, scheduling, reports — with a budget and tools per job and confirmation for anything irreversible.
A voice message is transcribed, and the answer can be spoken in Serbian — all on your infrastructure, the recording never goes to a third party.
A single point of access to models for your existing systems — the integration is done once — and migration preparation from legacy systems with a read-only device.
A layer in front of the model that cares how much gets sent at all: caching, batching, routing by task weight, a local model for sensitive work.
How the work is done
Irreversible steps — going live, changing data, switching off the old system — wait for your sign-off. Before and after handover.
What the system should do and where the data is. No forms, no offer up front — understanding the work comes first.
Who sends the request, who decides, which tools are allowed, how much may be spent and what is irreversible. That becomes a rule of the system, not a good intention.
A system blueprint and a work order with acceptance criteria: what is measured and when it is “done”. Building starts only after your sign-off.
On your infrastructure or, at your choice, in the cloud. Every user in their own space, keys in one place, a log of every step.
The criteria from the order are measured on real work, not promised. Whatever fails goes back for repair before handover.
The system is run long-term, not handed over and forgotten: model swaps, a new application on the same core, monitoring. Irreversible changes still require your sign-off.
House rules
The same rules apply in every system the house builds — from a personal assistant to an institution's portal.
Products and applications
All applications rest on the same core — SLAVKO, sovereign AI that remembers. Each card carries a maturity label.
Maturity labels: in production · in development · concept
Sovereign AI for organizations: persistent memory on your infrastructure, the model of choice — local or cloud — and answers checked before delivery. One core, many applications.
The product's main siteFor an individual or a small team: conversation by text or Serbian voice, it remembers what was agreed, keeps files and returns finished work — everyone in their own space.
More about the personal assistantAn operations portal that turns every request into a recorded job, with a one-click report for the authorities. AI prepares, a human decides.
Sovereign AI for the classroom: an assistant to the teacher and the school. Children's data stays in the country, and the system works even where the internet is unreliable.
More about the school assistantA private portal for a single building: notice board, one vote per apartment, a register of decisions with deadlines, repairs, documents and funds. Next phase: AI as an assistant to the building manager.
More about the portal and the AI managerClinical decision support for qualified healthcare professionals on the same core: it organizes references, guidelines and literature. It assists the physician, and the physician decides — a research prototype, not a medical device. The first phase is in production with a limited circle of physicians; the second phase, with more users, is in preparation.
More about ZDRAVKOA deadline tracker and error interpreter for the e-invoicing system, in Serbian, with answers only from the official text. It also runs on a local model, so financial data never leaves the company. The first phase is in production: the knowledge base from official sources; the answering layer and channels are the second phase, in preparation.
More about e-invoicingA single point of access for other systems, an explicit choice of model and a layer that lowers the cost — caching, batching, routing by task weight. The first phase is in production within our own systems; the commercial layer for other organizations is the second phase, in preparation.
A portable, local device that connects read-only to a legacy system, reads it and prepares a migration plan, so the engineer receives a finished analysis.
Physician-led health management and longevity: a record ready for the appointment, the right specialist at a known price, a guide through the administration and Zdravko as the patient's explainer — the decision is always the physician's. The first phase is in production with a small number of people; the second phase, with more users, is in preparation.
More about VO-ONEFounder
The house is led by Dr Ognjen Vučurević — a medical doctor and radiology resident who, since late 2024, has led the AI domain at an IT company in Belgrade: system architecture, specifications, acceptance criteria and the bridging of clinical, business and engineering teams.
Clinical discipline — differential diagnosis and the checks that must never be skipped — has been carried over into software architecture, where the same things are called validators, safety gates and the rule that a model's output is a candidate for a result.
The message runs opposite to the usual one: not “the system can do everything”, but “the system knows exactly where its limit is” — and that is why it is trusted where a mistake carries a cost.
Frequently asked questions
Answers to the questions asked most often before a first conversation.
A system designed around your specific business process, rather than a generic tool that the work adapts to. It knows who sends a request, who decides and what is irreversible, remembers your context tomorrow too, and is measured against acceptance criteria you signed off.
In local mode, no: the model runs on your infrastructure and the data stays where it was created. A cloud model is used only at your choice, with personal data removed when you require it.
The model's output is a candidate, not a result: it passes a check of structure, rules and an independent review before delivery. Whatever fails goes back for repair or is discarded, and an action that cannot be undone waits for a human to confirm.
They do not have to. Integration goes through a single point of access done once, and the assistant works in the channels you already use — web, Telegram, e-mail, voice. Migration from legacy systems is prepared with a device that only reads and changes nothing.
Any — local on your hardware or in the cloud. You choose explicitly, and the system runs exactly the chosen model; unavailability returns a clear error instead of a silent substitution. Swapping the model does not erase accumulated knowledge, because it is kept outside the model.
Metatron Production is the house that designs, builds and runs the systems; SLAVKO is its product — the core that remembers, with many applications. ZDRAVKO is the same core applied to healthcare, as support for the physician who decides; VO-ONE is physician-led health management for patients, in which Zdravko explains. Ulaz, the school assistant and the e-invoicing helper are applications on the same foundation.
With a conversation, by e-mail: what the system should do and where your data is. A process snapshot follows, then a blueprint with acceptance criteria — building starts only after your sign-off.
Contact
One message is enough: what the system should do and where your data is. No forms and no offer up front.
ognjen.vucurevic@gmail.comOur Slavko. Our knowledge.
One core, several public sites — each for its own topic.