AI-103 Practice Test

Answer choices

1. A. Utilize a template within your application code to dynamically construct the system prompt, incorporating variables that represent the desired persona and brand guidelines based on the specific customer interaction
2. B. Define the desired persona and brand guidelines in a single, lengthy system prompt that is included in every API request
3. C. Create a separate, detailed document outlining the persona and brand guidelines, and reference this document in the system prompt using a URL accessible by the Azure OpenAI service
4. D. Employ a combination of multiple, shorter system prompts, each focusing on a specific aspect of the desired persona and brand guidelines, and dynamically adjust their order based on the conversation context

Wrong-answer review

• B. Define the desired persona and brand guidelines in a single, lengthy system prompt that is included in every API request: A single long prompt is harder to adapt for different customer segments or interaction contexts.
• C. Create a separate, detailed document outlining the persona and brand guidelines, and reference this document in the system prompt using a URL accessible by the Azure OpenAI service: A URL reference in a prompt does not guarantee the model can retrieve or apply the external document.
• D. Employ a combination of multiple, shorter system prompts, each focusing on a specific aspect of the desired persona and brand guidelines, and dynamically adjust their order based on the conversation context: Multiple reordered prompts can make instruction priority and consistency harder to manage.

Answer choices

1. A. o1-mini, because it is a reasoning model capable of analyzing image structures
2. B. GPT-4o, because it is a multimodal model that accepts both text and image inputs natively
3. C. o1-preview, because it has vision capabilities enabled by default on Azure OpenAI
4. D. GPT-3.5-Turbo-Vision, because it is the standard cost-effective vision model

Answer choices

1. A. Configure an Azure ExpressRoute circuit and disable key authentication on the Azure OpenAI resource
2. B. Deploy an Azure API Management gateway in public subnet and restrict access using a client certificate
3. C. Create a Private Endpoint for the Azure OpenAI resource in your Virtual Network (VNet), and set the resource's firewall to disable public network access
4. D. Enable Service Tags on the Virtual Machine network security group (NSG) and set the Azure OpenAI inbound firewall to allow all traffic

Answer choices

1. A. Implement a custom Python function that reads PDFs from local disk using PyPDF2
2. B. Deploy Azure Cognitive Services Speech-to-Text to read the documents aloud to the agent
3. C. Configure the file search tool (vector store) in the agent definition and associate the uploaded document files with the agent thread
4. D. Configure a custom translator skillset on the database connection

Answer choices

1. A. The prebuilt Layout model, because it extracts text and determines the state of selection marks like checkboxes
2. B. The prebuilt Read model, because it extracts raw text and selection states
3. C. The prebuilt Receipt model, since it is optimized for small checkboxes
4. D. A custom classifier model trained in Azure Machine Learning to crop and analyze checkbox pixels

Answer choices

1. A. Modify the scoring profile to increase the weight of the vector scoring function and decrease the weight of the keyword scoring function
2. B. Configure a reranking filter that uses a cross-encoder model to score the relevance of each document based on the original user query and the document content, and then sorts the results by this reranking score
3. C. Implement a reranking filter in the query that prioritizes documents with a higher vector similarity score, regardless of keyword relevance
4. D. Adjust the indexing policy to include more keywords and phrases in the indexed documents to improve keyword-based search results

Answer choices

1. A. The prebuilt Read model, because it extracts table structures and paragraph reading order
2. B. The prebuilt Invoice model, since lease agreements resemble invoices in layout
3. C. The prebuilt Layout model, because it extracts text, selection marks, table structures, and document layout elements
4. D. A Custom Template model, trained on five sample lease agreements to define text boxes

Answer choices

1. A. Azure AI Vision with Custom Vision and OCR
2. B. Azure AI Vision with Computer Vision API and Form Recognizer
3. C. Azure AI Vision with Computer Vision API and Captioning API
4. D. Azure OpenAI Service with Visual Analysis capabilities

Answer choices

1. A. Prebuilt Layout model, because it classifies identity documents
2. B. Prebuilt Read model, because passport fonts are standardized for OCR
3. C. Prebuilt ID document model, because it is pre-trained to extract identity fields from passports and driver's licenses
4. D. Custom Template model, trained on five sample passports

Answer choices

1. A. Azure AI Vision with custom vision models trained on disease images
2. B. Azure OpenAI Service with Visual Analysis, leveraging its multimodal capabilities to understand image content and generate descriptive text
3. C. Azure AI Vision with OCR to extract text from labels and then use a custom logic app to determine the disease
4. D. Azure AI Vision with the Describe Images feature to generate captions and then manually review the results