MUNICH · PRAGUE · ON-PREMISE AI ENGINEERS

We build
AI Machines.

AI for hospitals, law firms and high-stakes data. Running on your hardware. Owned by you.

Live at Rowan Legal, T-Systems, Eurowag, University Hospital Pilsen, University Hospital Hradec Králové, Penta Hospitals.

Fakultní nemocnice Hradec Králové
Fakultní nemocnice Plzeň
Penta Hospitals
Nemocnice Náchod
Nemocnice Rychnov nad Kněžnou
Medicalc HIS
EUC — European Healthcare Network
Rowan Legal
Anywhere Legal
Die Netzwerkpartner
Eurowag
T-Systems
Heggel
RWE
ASO Systems
Fakultní nemocnice Hradec Králové
Fakultní nemocnice Plzeň
Penta Hospitals
Nemocnice Náchod
Nemocnice Rychnov nad Kněžnou
Medicalc HIS
EUC — European Healthcare Network
Rowan Legal
Anywhere Legal
Die Netzwerkpartner
Eurowag
T-Systems
Heggel
RWE
ASO Systems

INSIDE THE MACHINE

Six parts. One installed system.

Installed on your hardware. Yours to keep.

The Chassis

GPU servers in your building
No cloud routing
Air-gapped if needed

The Controls

API, web UI, native app
Fits into tools your team already uses

The Safety

Your data never touches our network
GDPR by architecture, not by audit

The Fuel

Voice recordings
Scanned documents
Legacy databases
Free-text records

The Engine

Transformers fine-tuned on your domain
Olingo model line for healthcare and legal
Not a generic API wrapper

The Connectivity

FHIR for hospitals
Dynamics + SharePoint for enterprise
Your existing systems, not new ones
Six parts. One installed system.

WHAT WE BELIEVE

Cloud AI is the wrong architecture for your data.

When the data is patient records, M&A drafts, or government correspondence, 'send it to the cloud' isn't a procurement decision — it's a liability. Every API call to a cloud LLM leaves a trace. Every query becomes training material. Or evidence in court. We built the alternative: AI that runs on your hardware, trained on your data, owned by you.

— Krystof Olik, Founder

No tokens. No quotas. No third-party access.

Your data never leaves your network. By architecture.

You own the model. You own the updates. You own the outcomes.

WHAT WE DO

Data translation, custom-trained for your domain.

We turn the messy data your business actually produces — scanned documents, PDFs, voice recordings, video, free-text notes, trading streams — into the structured records your downstream systems already understand. No hand-coded extractors, no generic API wrappers. Custom-trained transformer models, deployed on your hardware.

Unstructured input

Scanned documents
PDF
PDF reports
Voice recordings
Video footage
Free-text notes
Email threads
Custom-trained transformers

Structured output

01FHIR / HL7 medical records
02JSON for your APIs
03Tabular rows for your DB
04Domain-specific schemas
FHIR · HealthcareWhat a hospital's HIS receives back
{
  "resourceType": "Encounter",
  "patient": "Patient/PT-04821",
  "period": { "start": "2026-05-08T09:14" },
  "diagnosis": [
    { "code": "I10", "label": "Hypertension" }
  ]
}

Whatever the source. Whatever the target. Built for your domain — not a generic API.

VOICE → RECORDS

Doctors dictate. The structured record appears in your HIS.

Medical-grade transcription
OCR from scans & documents
Structured output schemas

DOCUMENTS → KNOWLEDGE

Decades of internal documents become an instant answer system.

RAG over your existing data
Legacy system integration
Real-time answers

FOR AI SOFTWARE COMPANIES

Your customers want it on their servers. We make that possible.

AI startups built for the cloud lose deals to buyers who can't accept cloud — hospitals, banks, law firms, government, defence. We port your stack to run on the customer's hardware, fully on-premise, so you can close those deals without rebuilding your product.

01

Port from cloud to on-premise

Your stack containerised and deployable on the customer's hardware. Air-gapped where required.

02

Replace API dependencies

Local model inference instead of OpenAI or Anthropic calls. Same capability, no external network.

03

Hardware sizing & operations

GPU specifications, deployment, and operational handoff to the customer's IT team.

FAQ

Frequently Asked Questions

Common questions about on-premise AI deployments

What does 'on-premise' mean in practice?

It means the AI model and all processing run on hardware located in your own data centre or server room. No data is sent to any external cloud service. You own the hardware, control the access, and the system can operate fully air-gapped — with no internet connection required.

How long does a deployment take?

Our fastest deployment was 3 weeks from contract to production. Typical complex integrations take 6–12 weeks depending on system complexity and client IT readiness.

Is private cloud an option if we're not ready for on-site deployment?

Yes. We offer private cloud deployments on EU-hosted infrastructure with full data isolation. This is a good starting point for teams moving towards on-premise ownership.

What happens after deployment?

We provide managed service support including model monitoring, performance optimisation, and updates. We can also transition to a fixed monthly maintenance model after the initial deployment.

How does pricing work?

Engagements are priced per project based on scope and complexity. We provide a fixed-price proposal after the 30-minute technical assessment. There are no token-based running costs — just infrastructure and support.

Contact

Talk to engineers, not sales.

We work with organisations ready to own their AI infrastructure. 30-minute session: we map your data flows and propose a deployment architecture. If we're not the right fit, we'll tell you.

WHAT HAPPENS NEXT

  1. 01We map your data flows and existing systems.
  2. 02We propose a deployment architecture for your infrastructure.
  3. 03You leave with a concrete plan. No pitch.
info@ollsoft.com

Trusted by hospitals, law firms, and industrial companies across Europe.