NAZARE
Manufacturing data infrastructure on three servers
Operate connections, ingestion, storage and queries together in an on-site cluster.
Explore the product
Products
Your factory’s know-how can stay inside.
Run equipment data and AI entirely on your infrastructure, with access and operating procedures designed for your site policies.
For owners & plant managers
Adopt AI while keeping production know-how inside.
For production, quality & maintenance
Use the information your work needs without uploading it externally.
The approach
How does it work?
- 01 Keep data inside
Fit sites with restrictions on data leaving
- 02 Set access
Use data within the scope of each task
- 03 Run internally
Start new uses within company policy
- 04 Review operations
Plan beyond the initial deployment
Details
Scope and considerations
Define location and operating ownership
NAZARE is an on-premises manufacturing platform using Kubernetes containers and distributed storage. Collection, storage, query and API modules support separate compute and storage expansion.
Isolated networks require plans for package import, updates, clocks, authentication, backups and support access. Define responsibilities between customer infrastructure teams and Ingkle.
Availability and deployment operations
Availability depends on node, disk, network, power and recovery design, not just server count. At least three nodes are recommended for operational clusters; a single node is for evaluation and demonstration.
Supported operating capabilities include containers, horizontal scaling, GitOps/IaC, rolling, canary and blue-green deployment, and resource monitoring. Validate recovery objectives against the actual design and contract.
Infrastructure review checklist
- Permitted paths between equipment and service networks
- Account provisioning, authentication and revocation
- External model APIs versus internal LLM serving
- Update import, verification and maintenance windows
- Backup location, restore tests and incident contacts
- Hardware procurement and maintenance ownership
AI compute is sized separately by model and concurrency; platform server guidance is not a universal LLM specification.
Start together
Tell us what you want to improve first.
Tell us about your equipment, data and priorities to review a path from connection to operations.