AI infrastructure

Local AI compute.

When deciding whether to use AI, businesses so far had to choose between benefit and data protection: public AI services are powerful, but your own data leaves the company. The Private AI Cloud makes that choice unnecessary - the same AI capability, on infrastructure that belongs exclusively to you or to us, in Switzerland.

Innocube: compact, lockable rack with network equipment, under a desk
The Innocube - compact enough to fit under a desk, lockable, barely audible.
Power consumption
An entire Innocube uses less power than a few old-fashioned light bulbs.
<60 W
on standby
200 - 300 W
under full load
Model quality
High answer quality thanks to a large language model.
250+ billion
parameters in the model
Response speed
Fast answers with hardly any waiting
<10 s
to first output (profile)
~30/s
tokens per second (writing speed)
Architecture
Ingesting documents never slows down answering questions.
2 separate engines
1 for ingestion, 1 for queries
Capabilities
What the Innocube can do

The Innocube is deliberately narrowly focused: a knowledge store. It takes in what is written and said in your company and answers questions about it in plain language - a company's entire knowledge, stored in a unit that fits under a desk. The technique behind it is called Retrieval-Augmented Generation (RAG): the platform first finds the relevant passages in your own documents and has the AI base its answer exclusively on them - instead of guessing from a general language model trained on the internet. Every answer stays tied to your company and traceable.

Document ingestion.

Reads in emails, voicemails, Zoom transcripts, Word, Excel, and PowerPoint files, and meeting notes, and prepares them for search - «ingestion» is the technical term for this reading-in and preparation.

Audio transcription.

Automatically turns recordings - voicemails, meetings, dictations - into text. The text is then searchable like any other document.

PDF conversion.

Converts ingested content to PDF - one uniform format for filing and display, regardless of the format it arrived in.

Metadata enrichment.

Automatically adds keywords and detected commitments (deliverables) to every document - extra information that makes finding things later more precise.

Semantic search & chat.

You talk to the AI about your documents and ask questions in plain language - instead of searching for keywords, you ask for meaning.

Integrations on request.

Connections to Microsoft Intune, SharePoint, Google Drive - and in principle any system with an interface (API). The Innocube runs either fully isolated or as local AI with an open door to your existing systems.

Flow - how the Innocube processes data
Sources
Emails
Voicemails & audio
Zoom transcripts
Word · Excel · PowerPoint
Meeting notes
etc.
Engine 1 · Ingestion
1
Transcription
Audio becomes text
2
PDF conversion
One uniform format for everything
3
Metadata enrichment
Keywords & deliverables
Storage
RAG datastore
Embedding + vector search: content is stored by meaning
Engine 2 · Query
Semantic search & chat
Questions in plain language to your documents
Three options, one concept

All three options run on the same base platform we developed. The difference is how much hardware belongs exclusively to you and where it sits.

OPTION 1 · SHARED
Innocube Shared Cloud

Your AI features run on shared hardware in our Swiss data center, alongside other Innodat clients - never on a third-party public AI service. The economical entry point without hardware of your own.

OPTION 2 · RENTED
Innocube Private Cloud

Your own hardware unit - compute node, firewall, networking, internet access - exclusively yours, operated on a rental basis in our data center. The hardware is sized to your team and desired model size; we maintain it and load new models as soon as they become available.

OPTION 3 · ON-SITE
Innocube at your site

The same hardware unit, purchased and installed at your site. Scope and terms are clarified in a personal conversation, matched to what you actually need.

Free initial consultation  →
INCLUDED WITH INNOCUBE
Innovault: your company knowledge, semantically searchable
🇨🇭 Swiss Data Sovereignty · GDPR & DSG compliant

Innovault ships with the Innocube. The application turns documents and recorded conversations into a knowledge base you can query in plain language. The answers are based on your own sources - stored and processed entirely in Switzerland on an Innocube, cloud or dedicated.

innovault.ch  →
Example

«What was agreed in the minutes of last month's meeting with client X?» or «In which contracts do we have a notice period under three months?» - the platform answers questions like these directly from your own documents, with a reference to the document the answer came from.

app.innovault.ch
innovault Chat Search documents, chats, projects… ⌘K LK L. Keller Personal Plus
LK Lukas Keller Personal Plus
Home
Library
Chat
Calendar
BY PROJECT
Contracts128
Finance86
Product54
🇨🇭 Swiss Data Sovereignty GDPR & DSG compliant
Supplier contracts - renewals Legal Finance
Which supplier contracts expire before December, and what notice periods apply?
Three agreements expire before December 2026. Müller Präzision AG ends on 30 September with a 60-day notice period 1 and Berger Logistik on 15 November with 90 days 2. Both renew automatically unless terminated in writing.
Draft a termination letter for Berger Logistik.
•••
Ask anything about your documents…
GROUNDED BY
1 PDF Rahmenvertrag_Mueller_2024.pdf
…kann mit einer Frist von 60 Tagen auf das Vertragsende gekündigt werden…
2 Berger_Logistik_SLA.docx
…Vertragsdauer bis 15.11.2026, Kündigungsfrist 90 Tage, Verlängerung automatisch…
Good afternoon, Lukas.
Your vault holds 214 documents.
Kickoff - Project Alpha 32 min · worked through
Scope for phase one agreed. Budget ceiling of CHF 180'000 confirmed.
3 deliverables 4 people 2 dates
What did I promise in March?
The revised cost estimate, due 20 March 1.
Home Library Chat Calendar
AI in your solution
What you can build with this infrastructure

A private knowledge platform, AI features in your FileMaker solution, operating models in detail.

See AI Solutions  →
How we work
AI in our own development process, too

How we use AI tools in our own development process - with clear rules about what goes to tools and what does not.

See AI Development  →
Questions about the Innocube? Contact us.
Contact us!