AI that works on your data.
We build AI agents and train language models for companies that can't send their data to a foreign cloud. The model is yours, it runs where your data lives, and every step is measured and logged.
Four ways to start
From a one-day answer to an agent in production. Every price is on the page.
An agent for one process
Reads mixed input, calls your systems, drafts or acts, and asks a named person where the rules say so. On your model, on your servers, with an activity log.
A model trained on your data
Fine-tuning and continued pre-training on your documents, measured on a frozen set from your cases. You get the weights.
AI Act review for SMEs
Which AI tools you use, what applies to each, an AI-literacy policy and disclosure texts. Delivered in five working days.
One day, one honest answer
Fifty of your real cases through several models. A written verdict on whether training or an agent pays off at all.
Working in Croatian, Serbian, a Baltic or another under-served language? Your language, your model →
Small steps, a number at every one
Evaluate
NDA first. Fifty real cases, several models, a frozen answer key. You see where AI helps and where it doesn't.
Pilot
One task, one model or agent, measured against the same set. Fixed price, agreed before we touch your data.
Deploy
On a laptop, a server in your rack or an air-gapped machine. Approvals and logging set up for your roles.
Keep it current
New data in, retrain, re-score, ship. When a better open base appears, we move to it within days.
We run our own agents first
Everything we offer is something we already depend on. Eight systems run on the same stack we would build for you.
What we learn, written down
Practical notes from our own work: regulation, training, evaluation. All posts →
Running a private LLM on your own server: models, setup and real costs
Which model size fits a narrow task, GGUF or vLLM, air-gapped setup, GDPR and the AI Act, cascades and EU cloud. What a private LLM on your server costs.
BlogFine-tuning or RAG for documents in your language?
When retrieval (RAG) is the right answer, when training a model on your data is, and when you need both. With numbers from our public work.
Compliance software in daily use
Two SaaS products for regulated work in Slovenia. They are why we know what an audit asks for.
Sensoram
Real-time temperature and humidity monitoring for pharmacies, hospitals, labs, food and logistics. EN 12830 reports, hash-chained records, escalating alarms.
sensoram.comEDC
Working-time records for Slovenian employers under ZEPDSV-A: web, offline mobile app, NFC and QR clock-in, automatic legal calculations, payroll export.
edc.pointer-it.siAlso from MediaAtlas: SemantiCord, EnergonX, AssetManiac, Mali Pokovci, Motion Capture Studio.
Bring us one process.
The job your team repeats most. We'll tell you, with numbers, whether AI should do it, and what it would take.