Create Pulumi Component for Azure Resource Layer to Support vLLM

[kkiani]

Description:

Following the discussion on integrating LLM/Foundation Model AI applications, this issue focuses on building a Pulumi component to define the necessary Azure resources for deploying a vLLM model using Damavand. The goal is to simplify infrastructure setup for developers building AI applications on Azure, using resources like Azure Synapse or a VM backend to host vLLM models.

Key Tasks:

• Implement the Pulumi component to provision Azure resources required for vLLM, such as VMs, storage accounts, and networking.
• Ensure the component follows best security practices.

Resources:

• Research Azure AI and other potential solutions for hosting vLLM models on Azure.
• Explore open-source tools or tutorials that support AI model hosting on Azure infrastructure.

Acceptance Criteria:

• A functional Pulumi component that provisions all necessary Azure infrastructure to support vLLM models.
• The component provide config args for parametrizing the commponent.
• The component should implement open/close principal by following the lazy loading pattern.
• Documentation in the methods and classes.