The operating system beneath AI applications.

The managed core for Agentic AI, on your infrastructure.

Explore the architecture

Selys

Your Apps

ISV Partner Apps

Managed AI services

Your Infrastructure

CPU

GPU

Storage

Network

What makes it an operating system?

At the AI platform level, Galene.AI OS coordinates resources, identity, execution and lifecycle management across tenants and applications. Each application consumes a controlled service environment instead of assembling its own foundation.

Managed by Galene.AI OS

Storage

Governed files, objects, databases and working volumes.

Keep evidence, state and results.

Persist source documents, knowledge indexes, application state and generated reports in governed storage.

CPU compute

Application services, background jobs and controlled execution sandboxes.

Run the work around the model.

Process documents, execute tools and analyse files in controlled environments, alongside your application's business logic.

GPU compute

Managed model serving and accelerated inference.

Put local models to work.

Generate responses, create embeddings and process multimodal inputs through assigned local model services.

Network

Managed endpoints, ingress and controlled outbound connectivity.

Connect the workflow to your systems.

Reach enterprise data, internal services and approved external providers within the configured network policy.

From infrastructure to applications.

Choose where the platform runs. Govern shared services, organise tenant access and build your business workflows.

Local. Hybrid. Air-gapped.

Keep inference local, combine it with approved external services, or operate in a supported disconnected environment.

Scalability: Grow capacity across validated configurations as applications and demand evolve.
Support & maintenance: Operate with agreed support, maintenance windows and deployment-specific access.
Managed updates: Maintain supported releases, including controlled offline updates for air-gapped deployments.
Cyber resilience: Combine configured access controls, security maintenance and planned backup and recovery.

More details

Scalability: Grow capacity across validated configurations as applications and demand evolve.
Support & maintenance: Operate with agreed support, maintenance windows and deployment-specific access.
Managed updates: Maintain supported releases, including controlled offline updates for air-gapped deployments.
Cyber resilience: Combine configured access controls, security maintenance and planned backup and recovery.

Govern and evolve.

Coordinate the platform lifecycle, shared resources and services assigned to each tenant.

Lifecycle & support: Coordinate supported updates, maintenance windows and the agreed support path.
Infrastructure & capacity: Set shared compute, storage and network boundaries for supported workloads.
Tenant provisioning: Create tenants and assign their capacity, resources and available services.
Models & shared services: Deploy supported local models, register approved external endpoints and allocate tenant services.

More details

Lifecycle & support: Coordinate supported updates, maintenance windows and the agreed support path.
Infrastructure & capacity: Set shared compute, storage and network boundaries for supported workloads.
Tenant provisioning: Create tenants and assign their capacity, resources and available services.
Models & shared services: Deploy supported local models, register approved external endpoints and allocate tenant services.

Configure your organisation.

Organise identities, data and applications within assigned services and permissions.

Identity & permissions: Connect your identity provider; manage users, roles and access with tenant-scoped RBAC.
Connectors & knowledge: Connect authorised sources; configure ingestion, provenance, knowledge collections and permitted data use.
Applications & access: Deploy supported apps, configure agents and assigned models, and grant users permitted access.
Help & support: Use guidance and tickets, escalating platform issues through the Control Plane's agreed support path.

More details

Identity & permissions: Connect your identity provider; manage users, roles and access with tenant-scoped RBAC.
Connectors & knowledge: Connect authorised sources; configure ingestion, provenance, knowledge collections and permitted data use.
Applications & access: Deploy supported apps, configure agents and assigned models, and grant users permitted access.
Help & support: Use guidance and tickets, escalating platform issues through the Control Plane's agreed support path.

Deliver the business workflow.

Define business logic and user experience using assigned services.

Business logic: Implement domain rules, decisions and the task sequence for your workflow.
User experience: Choose conversations, workspaces, forms or a dedicated interface for the task.
Service composition: Combine assigned inference, knowledge, tools and storage through supported interfaces.
Review & deliverables: Present evidence, check access, collect required approvals and deliver outputs to authorised users.

More details

Business logic: Implement domain rules, decisions and the task sequence for your workflow.
User experience: Choose conversations, workspaces, forms or a dedicated interface for the task.
Service composition: Combine assigned inference, knowledge, tools and storage through supported interfaces.
Review & deliverables: Present evidence, check access, collect required approvals and deliver outputs to authorised users.

Control Plane

Govern and evolve.

Coordinate the platform lifecycle, shared resources and services assigned to each tenant.

Lifecycle & support: Coordinate supported updates, maintenance windows and the agreed support path.
Infrastructure & capacity: Set shared compute, storage and network boundaries for supported workloads.
Tenant provisioning: Create tenants and assign their capacity, resources and available services.
Models & shared services: Deploy supported local models, register approved external endpoints and allocate tenant services.

More details

Lifecycle & support: Coordinate supported updates, maintenance windows and the agreed support path.
Infrastructure & capacity: Set shared compute, storage and network boundaries for supported workloads.
Tenant provisioning: Create tenants and assign their capacity, resources and available services.
Models & shared services: Deploy supported local models, register approved external endpoints and allocate tenant services.

Tenant Console

Configure your organisation.

Organise identities, data and applications within assigned services and permissions.

Identity & permissions: Connect your identity provider; manage users, roles and access with tenant-scoped RBAC.
Connectors & knowledge: Connect authorised sources; configure ingestion, provenance, knowledge collections and permitted data use.
Applications & access: Deploy supported apps, configure agents and assigned models, and grant users permitted access.
Help & support: Use guidance and tickets, escalating platform issues through the Control Plane's agreed support path.

More details

Identity & permissions: Connect your identity provider; manage users, roles and access with tenant-scoped RBAC.
Connectors & knowledge: Connect authorised sources; configure ingestion, provenance, knowledge collections and permitted data use.
Applications & access: Deploy supported apps, configure agents and assigned models, and grant users permitted access.
Help & support: Use guidance and tickets, escalating platform issues through the Control Plane's agreed support path.

Applications

Deliver the business workflow.

Define business logic and user experience using assigned services.

Business logic: Implement domain rules, decisions and the task sequence for your workflow.
User experience: Choose conversations, workspaces, forms or a dedicated interface for the task.
Service composition: Combine assigned inference, knowledge, tools and storage through supported interfaces.
Review & deliverables: Present evidence, check access, collect required approvals and deliver outputs to authorised users.

More details

Business logic: Implement domain rules, decisions and the task sequence for your workflow.
User experience: Choose conversations, workspaces, forms or a dedicated interface for the task.
Service composition: Combine assigned inference, knowledge, tools and storage through supported interfaces.
Review & deliverables: Present evidence, check access, collect required approvals and deliver outputs to authorised users.

Inside Galene.AI OS.

Build your application with reusable AI services, backed by a runtime we maintain.

Your applications

Reuse the services you need. Keep your own workflow and experience.

In-platform SDKUse platform services from your application code.

Service REST APIsCall services from your own applications and systems.

The compute, storage and operating environment beneath Galene.AI OS.

Hardware resources

Provide compute, storage and connectivity for platform services and applications, sized to the workload and validated deployment configuration.

What it includes

Single- and multi-node deployments
CPU and GPU compute
Internal and external storage
Configured network resources

In your application

Allocate CPU resources to background tasks, GPU capacity to inference and storage to application data and results.

Host OS & virtualisation

A host operating system and optional virtualisation environment sit beneath Galene.AI OS, with choices validated for each deployment.

What it includes

Red Hat Enterprise Linux
Rocky Linux
Ubuntu
DGX OS
OpenStack
Optional virtualisation

In your application

Select the base environment appropriate to your hardware and operating requirements before binding platform services to its resources.

Hardware resourcesSize the foundation to the workload.

Provide compute, storage and connectivity for platform services and applications, sized to the workload and validated deployment configuration.

What it includes

Single- and multi-node deployments
CPU and GPU compute
Internal and external storage
Configured network resources

In your application

Allocate CPU resources to background tasks, GPU capacity to inference and storage to application data and results.

OrchestrationCoordinate the platform's containerised workloads.

Coordinate containerised services using an orchestration option selected for the supported edition, release and operating environment.

What it includes

Kubernetes
Docker Compose
Podman

In your application

An appliance and a cluster use different orchestration patterns; the target deployment determines how services are installed and operated.

Host OS & virtualisationChoose the supported operating environment.

A host operating system and optional virtualisation environment sit beneath Galene.AI OS, with choices validated for each deployment.

What it includes

Red Hat Enterprise Linux
Rocky Linux
Ubuntu
DGX OS
OpenStack
Optional virtualisation

In your application

Select the base environment appropriate to your hardware and operating requirements before binding platform services to its resources.

Infrastructure

The compute, storage and operating environment beneath Galene.AI OS.

Hardware resources

Provide compute, storage and connectivity for platform services and applications, sized to the workload and validated deployment configuration.

What it includes

Single- and multi-node deployments
CPU and GPU compute
Internal and external storage
Configured network resources

In your application

Allocate CPU resources to background tasks, GPU capacity to inference and storage to application data and results.

Orchestration

Coordinate containerised services using an orchestration option selected for the supported edition, release and operating environment.

What it includes

Kubernetes
Docker Compose
Podman

In your application

An appliance and a cluster use different orchestration patterns; the target deployment determines how services are installed and operated.

Host OS & virtualisation

A host operating system and optional virtualisation environment sit beneath Galene.AI OS, with choices validated for each deployment.

What it includes

Red Hat Enterprise Linux
Rocky Linux
Ubuntu
DGX OS
OpenStack
Optional virtualisation

In your application

Select the base environment appropriate to your hardware and operating requirements before binding platform services to its resources.

Generative Nexus

Maintained runtime components power the services applications consume.

Observability

Collect logs, metrics, telemetry and service health information so operators can understand execution, investigate failures and assess resource pressure.

What it includes

Prometheus metrics
Grafana dashboards
Docker monitoring
Kubernetes monitoring
Logs and telemetry

In your application

Distinguish a slow model response from an ingestion failure or exhausted allocation before choosing which service needs attention.

GPU compute

Run assigned local inference services on available GPU capacity, covering text, vision, embeddings, speech and specialised model capabilities.

What it includes

LLM and VLM serving
Embedding models
Speech recognition (STT)
Speech synthesis (TTS)
Specialised models
Nexus Inference Servers

In your application

Run supported models against the GPU capacity available to your application, according to its inference requirements.

Storage

Supply file, object, database, cache and working-volume services for application data, knowledge indexes, state and generated outputs.

What it includes

Nexus VFS
RustFS
Redis
PostgreSQL
pgvector vector search
Apache AGE graph queries
Scratch persistent volumes
Container registry

In your application

Keep application data in persistent storage and temporary files in working volumes, with defined retention and recovery policies.

CPU compute

Run application services, queued work and controlled tool execution through maintained controllers, sandboxes and browser components beneath the service interfaces.

What it includes

Nexus Application Controller
Temporal queues
Nexus sandboxes
Chromium

In your application

Run isolated application tasks while background jobs process data and save results to managed storage.

Security & identity

Protect service access with secrets, scoped identities and encrypted transport, using the maintained identity and key-management components of the reference stack.

What it includes

Nexus Vault
Multi-factor authentication
Key management (KMS)
Keycloak identity provider
Istio mutual TLS mesh
Ingress

In your application

A connector uses scoped credentials for its approved source while service-to-service communication follows the configured encrypted transport boundary.

Generative Pillars

Shared services turn platform resources into application capabilities.

Multitenancy

Give each organisation scoped resources, encryption, quotas and administration, with backup and lifecycle controls appropriate to the configured deployment.

What it includes

Enterprise data separation
Per-tenant encryption
Storage, inference and compute quotas
Backup and recovery capabilities
Supported platform updates
Tenant-scoped resources and metrics

In your application

Organise application access and service limits by tenant, keeping knowledge, identities and operational contexts separate.

Authentication & authorisation

Authenticate users through connected identity providers and apply roles, groups and permissions to applications, knowledge sources and service operations.

What it includes

SSO, LDAP and SAML2
Multi-factor authentication
Role-based access control (RBAC)
Microsoft and Google OAuth 2.1
IAM permissions
Tenant-scoped users and groups

In your application

Give users access to authorised application features and data without granting tenant administration rights or access to unrelated resources.

Harness & agents

Coordinate context, inference, tools and task progression so agents can perform multi-step work using shared memory, scheduling and model-routing services.

What it includes

Agentic loop and MCP clients
Skills, tools and Skills Builder
Memory and context compaction
Scheduling and long-running goals
Model router

In your application

Let agents use authorised tools and sources across multiple steps while retaining the context required by your application workflow.

Data ingestion & search

Connect authorised sources, process multimodal content and build retrieval structures, preserving the source and retrieval permissions required by each application.

What it includes

RAG, GraphRAG and VLRAG
Multimodal ingestion pipelines
SQL and NoSQL connectors
NFS connectors
Microsoft and Google Drive connectors
Embeddings and incremental synchronisation

In your application

Retrieve relevant content and relationships from authorised sources to ground model responses in the knowledge available to your application.

Web providers

Supply controlled web access and customer-key providers for workflows that need external information, within the deployment's approved connectivity and data boundary.

What it includes

Agent web fetch
Agent web search
Agent web browsing
BYOK search and retrieval providers

In your application

Give connected applications access to approved web sources; use permitted local knowledge in air-gapped configurations.

Application catalogue

Distribute versioned application packages with declared requirements, access roles and resource dependencies, then deploy them within available tenant services and capacity.

What it includes

Versioned declarative manifests
Automated deployment and registry management
Infrastructure-as-code workflows
Per-application roles and permissions
Per-application usage metrics
Software-defined storage, databases and queues

In your application

A custom application declares its storage, database and queue requirements so a versioned package can be deployed against the target tenant.

Help Center

Bring operational guidance, internal ticketing and the defined maintenance channel into the environment so tenant teams and operators can resolve issues.

What it includes

Interactive knowledge base
Integrated internal ticketing
Galene.AI support and maintenance channel

In your application

A tenant administrator consults the knowledge base about a connector issue, then routes an unresolved incident through the agreed support path.

Model inference

Request inference through assigned model services, combining local serving and approved external providers with quotas, routing and fallback inside the intended boundary.

What it includes

Hybrid gateway for local and cloud models
BYOK external providers
Inference quotas and limits
Local model catalogue and deployment
Speech models for STT and TTS
Assigned slots and bounded fallback

In your application

Choose supported local inference or approved external providers according to each application's data boundary and model requirements.

Compute sandboxes

Execute code and skills in bounded working environments, with scratch volumes, role-based resource separation and configurable network policy.

What it includes

Persistent scratch volumes
Skill, code and Bash execution
Network modes for none, full or restricted access
Allow/deny rules and technical domains
RBAC resource separation
Resource limits and permissions

In your application

Execute application code with temporary files and approved tools, within configured permissions, network rules and resource limits.

Generative Shield

Shared AI controls at configured service and runtime integration points.

Guardrailing

Apply configured guardrails to inputs and outputs, informed by model risk and intended use at supported service and runtime integration points.

What it includes

Risk-based model catalogue
AI Act-informed risk classification
Bidirectional runtime checks
Defences against adversarial use

In your application

Inspect a request and its generated response against configured policies before the application presents the result to authorised users.

AI watermarking

Add inference-time marking and private detection for supported model and output configurations, with suitability assessed for the application's actual deployment.

What it includes

Tenant- and model-level watermarking
Real-time marking at inference
Bayesian private watermark detection

In your application

Assess whether a selected model can mark its generated outputs and whether the configured detector can identify those marks.

AI governance

Maintain intended-use records and model reports to support risk assessment, alongside the application's business rules, approval steps and legal responsibilities.

What it includes

Intended-purpose repository and history
Curated model catalogue reports
AI Act risk assessment support

In your application

Review a model's documented intended purpose and assessment information before approving its role in an application's business process.

AI cybersecurity

Detect jailbreak and prompt-injection patterns at supported runtime integration points, complementing platform identity controls, infrastructure security and application-specific checks.

What it includes

Real-time jailbreak detection
Real-time prompt-injection detection
Inspection of supported runtime paths

In your application

Inspect untrusted retrieved content before an agent acts on embedded instructions, within the configured protection path.

APIs across the stack.

Administer the platform, consume tenant services and expose application logic.

Control Plane APIs

Provision tenants, assign capacity and deploy applications and supported models.

Tenant service APIs

Use inference, agents, knowledge and web services within tenant permissions and quotas.

Application APIs

Connect business systems to your application's own workflows and authorisation rules.

API

Call native REST interfaces, with OpenAI compatibility for supported model services.

SDK

Reuse supported service interfaces from your application code.

Galene.AI Faber

Give compatible coding agents the platform’s service contracts and development conventions.

One platform. Three ways to deploy.

A compact appliance, an enterprise cluster or your existing infrastructure. Choose the footprint that fits the way you work.

Elettra

Compact. Simple. Turnkey.

A compact AI environment on a single node, sized for a focused team or workload.

Dell Pro Max with GB10 compact workstation

Docker-based
Bare metal or VM
Single node

Capabilities

Compact compute: Standard Dell OEM workstation configurations, or installation on GPU-enabled virtual machines.
Operational visibility: Container-based observability for the services running on your node.
Tenant data protection: Per-tenant encryption at rest.
Simple installation: Debian package installer with Ubuntu and DGX OS support.

Hardware & GPUs

Hardware options

Dell Pro Max GB10
Dell Pro Max Tower T2

NVIDIA GPU options

GB10 Superchip
RTX PRO 6000

Arethusa

Scalable. Resilient. Enterprise-grade.

A multi-node foundation for shared AI services, growing capacity and enterprise operations.

OpenStack + K8s
Bare metal
Multi-node

Air-gapped option

Capabilities

Cluster resilience: Standard Dell OEM rack configurations with multiple inference and storage nodes, designed for enterprise resilience.
Protected workloads: Per-tenant encryption at rest and in transit with mTLS. Enforced CPU and RAM limits for deployed applications.
Connected operations: Enterprise infrastructure observability and integration with external centralised block and object storage.
Enterprise hosts: Red Hat Enterprise Linux and Rocky Linux support.

Hardware & GPUs

Most deployed servers

Dell PowerEdge R770
Dell PowerEdge R7725
Dell PowerEdge XE7745
Dell PowerEdge XE9785

Other server configurations can be validated for your deployment.

NVIDIA GPU options

RTX PRO 6000
H100
H200
B200
B300

Castalia

Built around your infrastructure.

Run on your infrastructure with the modules you need.

Your infrastructure

Compute

Network

Storage

Existing Kubernetes
Custom integrations

Air-gapped option

Capabilities

Your infrastructure: Custom compute, network and storage configurations shaped around your requirements.
Your Kubernetes: Install on an existing cluster, including third-party managed Kubernetes environments.
Modular integration: Galene.AI modules can be provided and deployed independently, with custom infrastructure integration.
Tailored scope: A custom pricing model aligned with the deployment and modules you need.

Hardware & GPUs

Hardware options

Custom enterprise compute, network and storage

NVIDIA GPU options

RTX PRO 6000
H100
H200
B200
B300

Find the right partner for your deployment.

Connect with partners to plan, integrate and deploy Galene.AI OS on your infrastructure.

Find a partner