AI infrastructure setup
When data sensitivity or regulatory requirements rule out cloud services, we set up the AI environment on the company's own infrastructure, from procuring the equipment to a working system.
When data sensitivity or regulatory requirements rule out cloud services, we set up the AI environment on the company's own infrastructure, from procuring the equipment to a working system.
Uses, data volume and requirements.
Selecting and procuring the hardware.
Models, runtime environment and user interface.
Testing, documentation and training.
Below we describe when on-premise operation is justified, what the setup includes, and how the work proceeds from assessment to handover.
On-premise infrastructure is needed when the processed data must not leave the organisation. This can include personal and health data, technical documentation, contracts or financial data, as well as sectors where regulation or customer requirements demand that data is handled in-house.
With high and steady usage, running on-premise can also be more cost-effective in the long term than a usage-based cloud service.
We assess the expected uses, the number of users and the data volume, and on that basis determine the computing capacity required and the appropriate models.
We help select and procure the servers and graphics accelerators, in line with the existing IT environment.
We install and configure the open-source models and the runtime environment, set up the user interface and access management, and connect the system to the internal data sources required.
The system runs inside the company network, access is tied to roles, and usage can be logged.
Not every task has to run on-premise. A common approach is to process sensitive data on the company's own infrastructure and use a cloud service for other tasks. We draw this line during the assessment.
After handover we can take on monitoring the system, updating the models and expanding capacity. If the company's own IT team operates the system, we provide documentation and training.
It depends on the chosen model and the number of users. Based on the assessment we recommend the servers and graphics accelerators.
It depends on the chosen model and the number of users. Based on the assessment we recommend the servers and graphics accelerators.