SYS-DEP / Preliminary Specification

On-Premise Deployment

Air-gapped deployment.
Your infrastructure, your control.

For organizations where cloud deployment — even EU-hosted, VPC-isolated — is insufficient. evidion runs entirely within your network perimeter. No external API calls. No data egress. No exceptions.

Preliminary Specification

On-premise deployment is really complex and we have this roadmap outlined if demand is strong. Due to high CapEx (hardware) and maintenance overhead, we generally recommend our VPC-Isolated Cloud — it offers equivalent data sovereignty with 85% lower TCO and instant updates.

No external inference

Model processing runs on your hardware. There are no outbound calls to evidion or third-party providers.

Your IT controls updates

Shipped as a containerized app. Your team pulls updates from your private registry. No remote access.

Data stays inside

FHA inputs, drafts, and system descriptions are processed and stored locally. Nothing is transmitted.

Full local audit

All activity logs and audit trails are stored in your own PostgreSQL instance. You own the data.

Technical Stack

Infrastructure Requirements

ComponentRequirementNotes
Container runtimeDocker 24+Kubernetes recommended for multi-user
DatabasePostgreSQL 15+Structured data, audit trail, user management
Object storageMinIO or S3-compatiblePDF storage, document versioning
Container registryPrivate (Harbor or equivalent)Required for receiving updates
GPUNVIDIA L40S or H100See hardware tiers below
NetworkInternal LAN / VLANAir-gapped configuration recommended

No cloud dependencies in the operational path. Internet access is only required during initial deployment and when your team chooses to pull updates.

Reference Architecture

Hardware Tiers

Hardware is customer-provided. The estimates below are reference points for scoping conversations — not quotes. If you have existing GPU infrastructure, we assess compatibility before any commercial discussion.

TIER 1 · Safety Lab DEP-L

Estimated hardware CapEx: from €15,000
Entry Level
1× NVIDIA L40S 48 GB
64 GB RAM · 2 TB SSD
128k token context

Suitable for smaller teams running focused subsystem assessments. Not recommended for full PSSA/SSA workflows across complex architectures.

TIER 2 · Enterprise Safety Cluster DEP-E

Estimated hardware CapEx: from €90,000
Recommended
2× NVIDIA H100 80 GB
128 GB RAM · 4 TB NVMe
1M token context

Recommended for full ARP4761 and ISO 26262 workflows. Handles complete PSSA/SSA documents in 10–30 minutes. Single team session support.

TIER 3 · High-Performance Isolated Cluster DEP-H

Estimated hardware CapEx: from €200,000
High Performance
4× NVIDIA H100 80 GB HGX
256 GB RAM · 8 TB NVMe RAID
Multi-user concurrent sessions

For full air-gapped, multi-team deployment across complete aircraft system portfolios. Defense and top-tier OEM use cases only.

Adoption Roadmap
1

Infrastructure validation call

We review your existing hardware and network. 30 minutes with your IT architect. No commitment at this stage.

2

Deployment scoping

We recommend a tier and define the architecture: network isolation, storage, and update mechanism.

3

Commercial terms

License scoped per engagement based on team size and tier. Full documentation for procurement provided.

4

Guided deployment

Your team deploys from our documentation. We support the initial setup. Post-handover, you own all updates.

Join the pilot programme.

We are scoping the first on-premise deployments now. No commitment required to have the conversation.

We respond within 48 hours. No automated sequences.
On-premise deployment is currently in pilot.
Prefer a direct conversation? contact@evidion.dev