On-prem LLM
Private AI — LLMs in your environment, not on a public cloud
Use large language models while keeping data, IP and PII inside a controlled on-prem or private-cloud boundary, with an AI Gateway.
Who it is for
Organisations with sensitive documents, internal procedures or rules that forbid sending text to public LLM APIs.
The problem
Public AI tools treat context leakage as acceptable. For contracts, drafts and customer data it is not.
What is included
- Hardware delivery and install where needed
- Model quantisation and fine-tuning
- Custom apps and agents
- AI Gateway: who may send what to the model
What is not included
- Training on your data without a written scope and retention rules
- Public multi-tenant LLM as the only path
How implementation looks
Use-case and data classification first, then model and runtime, then gateway and agents. Portal: ai.konekt.rs.
Integrations
Documents, helpdesk, CRM and wiki — only through approved connectors and policies.
Security and operations
Service isolation, strict access, no tenant mixing. The goal is to remove leakage and unauthorised use.
Support
Model and gateway operations can sit on Managed IT. Prompt and agent changes stay on an iteration backlog.
Questions from sales conversations
Does data go to ChatGPT?
Not in this model. Runtime is on-prem or a private cloud you control.
Do we need a GPU cluster?
It depends on the model and concurrency. A smaller on-prem node or shared private capacity can be enough.
What is the AI Gateway?
The control point: auth, rate limits, data filters and an audit trail of model calls.
Next step