Deployment Options
Our AI-powered document intelligence platform can be deployed in three configurations — Cloud, Hybrid, and Full On-Premises — to meet your firm's data-security and compliance requirements.
The Hybrid modelRECOMMENDED is ideal for engineering and infrastructure organisations. It keeps all source data within your network, while leveraging secure Australian cloud compute for AI processing.
This approach delivers full performance and full compliance — without the cost of full on-prem hardware.
DEPLOYMENT
Choose Your Configuration
Three deployment models designed to balance security, performance, and cost. Each configuration meets different compliance and infrastructure requirements.
| Feature | Cloud | HybridRECOMMENDED | Full On-Premises |
|---|---|---|---|
| Where documents live | All files and processing in the cloud | Files stay on your servers; only transient text snippets processed in cloud | All data and compute remain on your servers |
| Security ownership | Cloud provider | Shared — you keep custody of all data; cloud handles transient compute | 100% client-managed |
| Performance | High, bandwidth-dependent | High — local file access + cloud GPU compute | Moderate (limited to local CPU/GPU) |
| Setup & cost | Low | Medium (light local agent + cloud embedding) | High (hardware + maintenance) |
| Maintenance | Provider-managed | Auto-updates in cloud; minimal local support | Fully client-managed |
| Compliance (ISO, NATA, AU data-residency) | May require extra controls | ✓Fully compliant | ✓Fully compliant |
| AI model updates | Automatic | ✓Automatic | Manual |
HYBRID ARCHITECTURE
How the Hybrid Model Works
1Local Connector – Secure Bridge
A lightweight service installed on your own server or VM:
- •Reads and parses approved files locally (PDF, Word, CAD, drawings)
- •Extracts text and metadata for embedding
- •Sends only small text fragments (hundreds of words) over encrypted outbound HTTPS to Azure for embedding
- •Receives numeric vectors, discards text, and stores nothing readable
- •Fetches original text locally when users search, ensuring data never leaves your network
2Cloud Embedding & Vector Storage
- •Uses secure cloud infrastructure (e.g., Azure Australia East) for transient embedding — text is processed in memory only
- •Produces numeric vectors stored in a secure, region-locked vector database
- •Stores only minimal metadata (doc ID, page #, project code)
- •No readable content or files retained in the cloud
Your Data Stays Local
Project files never leave your network
Mathematical Representations Only
Only vectors are sent to secure cloud model
No Raw Content Exposed
No text, drawings, or metadata are exposed
Full AI Capability
Without managing expensive GPU hardware
3Query & Retrieval
User submits a search in the web app
Hosted securely in the cloud
Query is embedded and matched against vectors
AI processing finds relevant content
Connector retrieves matching snippets locally
Full source traceability and context
Your data stays secure
No full documents or sensitive data leave your network
DATA SECURITY
Metadata Handling
| Type | Example | Stored In |
|---|---|---|
| Non-sensitive | doc_id, page_number, keywords | ☁️ Cloud vector DB |
| Sensitive | file path, author, client name, project title | 🏢 Local metadata index |
| Vectors | numeric embeddings [0.254, -0.088, …] | ☁️ Cloud |
| Original files | PDFs, Word, DWG | 🏢 Local storage |
Result: The cloud knows what's relevant, but only your system can reconstruct the text.
TECHNICAL OVERVIEW
Data Flow & Security
Ready to Learn More?
Schedule a call to discuss how the Hybrid deployment model can work for your firm.