See the architecture behind the broader platform
Explore the planned layers, infrastructure roles, and operating controls for the expanded Phison AI Data Platform. The broader platform will support more applications and multiple servers. For the first AIDP offering, see the current single-server GenBI solution.
How the layers will connect
Applications use model services and business context to complete tasks. Resource orchestration manages the compute, memory, and storage those services need. Each layer has a defined role; the connections depend on the application and workload.
Agent platform
Build, govern, and scale enterprise agents.
- Agentic hub
- AI customer service
- Document intelligence
- Enterprise knowledge
AI data intelligence layer
Powered by Wren AI. Connect business data and definitions to conversational analytics.
- Data integration
- Databases and data sources
- Generative BI
Model and AIOps
Deploy, operate, and optimize supported AI models.
- MLOps / LLMOps
- Model management
- Vision
- Voice
- Reasoning
Resource orchestration layer
Manage compute, memory,
and storage across supported infrastructure.
- Infrastructure management
- Workspace and access
- Operations and monitoring
Hardware layer
GPU compute, Pascari aiDAPTIV™ cache memory and middleware,
and Pascari enterprise SSDs.
- GPU servers with aiDAPTIV
- Cache servers aiDAPTIV SSD
- Storage servers with Pascari SSD
Hardware configurations for the workload
The first GenBI offering uses current reference configurations from four or eight GPUs in a single server. The broader platform will add supported multi-server configurations and additional infrastructure roles.
We’ll help you choose the model and configuration for your business needs. Hardware, software compatibility, and validation results for the expanded platform will be defined by release.
Infrastructure role*
Planned purpose
* Planned infrastructure roles. They do not define the components included with the current single-server GenBI offer.
Scalable to meet your growing needs
Within a compatible GenBI server, GPU upgrades depend on available slots, power, cooling, and GPU support. The broader platform will add a separate path across servers.
Security, governance, and operations features
The broader platform roadmap includes controls for infrastructure access, shared resources, and operations. The supported controls and configuration requirements will be defined for each release.
Security and access control
Role-based access and permission controls for supported compute, storage, and services, coordinated with application access.
Multi-tenant isolation
Role-based access and permission controls for supported compute, storage, and services, coordinated with application access.
Audit logging
Activity records for investigating operations and reviewing access.
Governance
Policies and approval controls for supported AI operations, alongside application-specific business definitions and answer validation.
High availability
Health monitoring and supported failover options for multi-server deployments. Recovery behavior will depend on application state, redundancy, and deployment design.
Monitoring and maintenance
Visibility into performance, utilization, cost, and system health. Maintenance procedures will reflect the requirements of the supported configuration.
One coordinated team behind your deployment
Phison, application partners, delivery partners, and your team each contribute to deployment. Agree on ownership for infrastructure, application integration, data access, validation, and support before implementation begins.
Role
Contribution
Illustrative delivery model. Responsibilities and support terms are agreed upon for each engagement.
Dive deeper
Explore the current GenBI solution and the planned broader platform. Review technical resources as they become available or bring your requirements to a discussion with Phison.
Bring your technical requirements
Discuss the broader platform coming soon, including deployment, integration, planned applications, and multi-server operations.