AI-900 Practice Question: Describe features of generative AI workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  A low-level kernel module that optimises GPU utilisation for Azure OpenAI inference

  Why it's wrong here

  Semantic Kernel is not a low-level kernel module at all; the word 'kernel' here refers to the SDK's central orchestration engine, which coordinates AI services, not to an operating-system or GPU driver component. Low-level GPU kernel optimization is a hardware/compiler concern (e.g., CUDA or ROCm), whereas Semantic Kernel is a high-level .NET/Python/Java library that abstracts LLM calls.

• ✓

  Microsoft's open-source SDK for orchestrating LLMs with plugins, memory, and planning

  Why this is correct

  Semantic Kernel is Microsoft's open-source SDK (available on GitHub, under MIT license) that lets developers orchestrate large language models using plugins (also called skills), semantic/vector memory (to store and retrieve embeddings), and AI planners that break a user's request into a sequence of function calls. It provides connectors to Azure OpenAI, OpenAI, and other LLMs, along with a kernel object that manages the context, memory, and prompt templates for building complex, agent-like applications.

• ✗

  A tool for evaluating the semantic accuracy of Azure OpenAI model responses

  Why it's wrong here

  Semantic Kernel is not an evaluation or benchmarking tool; it is a framework for composing applications, not for measuring response quality. While you could build an evaluator on top of it, the SDK itself does not compute semantic accuracy, similarity scores, or faithfulness metrics—that would be the job of services like Azure AI Evaluation or prompt-flow's built-in evaluators, which are separate from Semantic Kernel's orchestration purpose.

• ✗

  Microsoft's proprietary alternative to Azure OpenAI for internal use only

  Why it's wrong here

  Semantic Kernel is neither proprietary nor an Azure OpenAI replacement; it is an open-source, cross-platform SDK (MIT license) that deliberately works alongside Azure OpenAI, OpenAI, and even other local or remote LLMs. Rather than being an alternative inference service, it acts as a coordination layer that injects context, manages memory, and calls models dynamically, which is the opposite of an internal-only Microsoft tool.

Go deeper

Related to this question