HOTSPOT (Drag and Drop is not supported)You are developing an application that converts text into spoken audio and saves the synthesized audio to a file by using Azure Speech in Foundry Tools.How should you complete the Python code? To answer, select the appropriate option in the answer area.Note: Each correct selection is worth one point.Hot Area:
Answer(s): A
Box: AudioOutputConfig(filename="output.wav")To complete the code and save the synthesized audio directly to a file, use the code AudioOutputConfig (filename="path_to_file.wav").Class Used: The AudioOutputConfig class is explicitly used to manage speech synthesis destination routing (such as physical files, speakers, or digital streams).File Format: The Azure Speech SDK natively saves output audio files generated through this configuration in the WAV format.
https://stackoverflow.com/questions/77923835/how-to-save-a-stream-object-in-azure-text-to-speech-without-speaking-the-text-us
HOTSPOT (Drag and Drop is not supported)Select the answer that correctly completes the sentence.Hot Area:
Box: system prompt To define an agent’s role and behaviors, you must configure a ________ for the agent.Configuring a system prompt (or system message / instructions) is the primary way to define an agent's role, persona, scope, and operational behaviors in Azure AI.System prompts act as the agent's "job description". They dictate how the underlying model interprets intent and executes tasks before it ever sees a user's input.
https://learn.microsoft.com/en-us/azure/microsoft-discovery/how-to-prompt-engineering
HOTSPOT (Drag and Drop is not supported)You are developing an application that analyze invoices by using Azure Content Understanding in FoundryTools.You need to ensure that the application retrieves the analysis results after processing completes.How should you complete the Python code? To answer, select the appropriate option in the answer area.Note: Each correct selection is worth one point.Hot Area:
Box: result In the official Azure SDKs for Python, asynchronous or long-running service operations (such as document analysis) return a poller object (an instance of AzureOperationPoller).Calling the .result() method on this poller tells the application to wait (block execution) until the document extraction processing completely finishes.Once completed, it retrieves and returns the structured analysis results (such as extracted fields or tables).
https://learn.microsoft.com/de-at/python/api/azure-ai-contentunderstanding/azure.ai.contentunderstanding
DRAG DROP (Drag and Drop is not supported)You have a Microsoft Foundry project named project1 that contains an Azure OpenAI resource named Resource1.To Resource1, you deploy a gpt-4.1-mini model by using a model deployment named my-mini-gpt.You need to connect to my-mini-gpt from an application.How should you complete the Python code? To answer, drag the appropriate values to the correct targets. Each value may be used once, more than once, or not at all.Note: Each correct selection is worth one point.Select and Place:
Box 1: resource1 Resource name: Azure OpenAI endpoints follow the specific standard structure of https://<resource-name>://. This routes your API requests directly to the dedicated Azure OpenAI resource containing your deployments.Box 2: my-mini-gpt Name of the custom model deployment: When using the OpenAI SDK with Azure OpenAI, the model parameter expects your unique, user-defined deployment name rather than the underlying base model name (like gpt-4.1-mini).Incorrect: Project name: Microsoft Azure AI Foundry projects manage connections to resources, but the actual network endpoint belongs specifically to the underlying Azure OpenAI service resource.The name of the custom model / just gpt-4.1-mini: Passing the base model name or model definition string will result in a 404 deployment not found error. Azure OpenAI relies strictly on the deployment identifier to look up your hosted model instance.
https://docs.weaviate.io/weaviate/model-providers/openai-azure/generative
HOTSPOT (Drag and Drop is not supported)For each of the following statements, select Yes if the statement is true. Otherwise, select No.Note: Each correct selection is worth one point.Hot Area:
Box 1: No No - An AI generative model is retrained before performing each user request.Generative models in Azure AI are not retrained before each user request. Retraining is a computationally heavy and time-consuming process. Instead, the service utilizes powerful, pre-trained foundation models that process your prompts in real-time.Box 2: No No- An AI agent responds by copying and pasting answers stored in a database.AI agents do not simply copy and paste pre-written answers from a database. Instead, they use a generative approach. When a user asks a question, the agent retrieves the relevant context, understands the intent, and dynamically writes a new, conversational response—often citing its sources.Box 3: Yes Yes -An AI agent uses a generative AI model to establish actions based on user input.Within Azure AI—specifically through services like the Microsoft Foundry Agent Service and Azure OpenAI Service—an AI agent uses a underlying generative AI model as its "reasoning engine". Instead of relying on rigid, pre-programmed code, it interprets natural language user input to dynamically plan, orchestrate workflows, and choose the most appropriate actions.
https://learn.microsoft.com/en-us/azure/foundry-classic/openai/faq
DRAG DROP (Drag and Drop is not supported)You are reviewing best practices for using AI at your company.Which Microsoft responsible AI principle is each task an example of? To answer, drag the appropriate principles to the correct tasks. Each task may be used once, more than once, or not at all. You may need to drag the split bar between panes or scroll to view content.Note: Each correct match is worth one point.Select and Place:
Box 1: Fairness Evaluating model outputs to ensure that decisions are not biased against specific demographic groups.Evaluating model outputs to ensure decisions are not biased against specific demographic groups is a direct example of fairness.In the Microsoft Azure AI principles, fairness means AI systems should treat all people equitably and avoid harming specific, protected groups of individuals.Box 2: Privacy and Security Encrypting sensitive customer data and restricting system access to authorized personnel.These practices align directly with the Privacy and Security principle of Microsoft's Azure AI framework.Data Protection: Customer data must be protected at rest and in transit using industry-standard encryption protocols.Access Control: Systems must use Role-Based Access Control (RBAC) and the principle of least privilege to restrict access.Compliance: AI systems must adhere to strict regulatory frameworks (like GDPR or HIPAA) regarding data sovereignty and user privacy.Box 3: Transparency Informing users when they are interacting with an AI system and explaining the system's capabilities and limitations.Under Microsoft's Azure AI principles, transparency ensures that stakeholders understand how AI systems are built, run, and integrated.User Awareness: People have a right to know if they are interacting with a human or a machine.Expectation Management: Clearly stating capabilities prevents users from over-relying on the system.Risk Mitigation: Explaining limitations helps users catch errors, biases, or "hallucinations."Informed Consent: Users can decide whether they want to proceed based on how the system works.Box 4: Reliability and Safety Testing AI systems under different conditions to reduce unexpected failures.Testing AI systems under different conditions to reduce unexpected failure is a direct application of the Reliability and Safety principle.Consistent Operation: Systems must perform reliably under daily conditions and unexpected situations.Risk Mitigation: Rigorous testing anticipates edge cases to prevent physical, financial, or psychological harm.Error Handling: It ensures the system can fail gracefully without causing systemic damage.Split-Condition Testing: Evaluating models against diverse datasets simulates real-world unpredictability.
https://www.edureka.co/blog/ai-on-microsoft-azure
Box 1: Yes Yes - Generating a response to a user prompt occurs during the inference stage.What Inference Means: In Azure AI and broader machine learning, inference is the live execution phase. It happens when a fully trained model processes new, unseen inputs (your prompt) to generate predictions or content (the response). The Opposite Phase: This stands in contrast to the training phase, where the model learns patterns, weights, and language rules from massive historical datasets. Box 2: No No - A generative AI model generates responses by copying stored documents directly from the model’s training data.Generative AI does not store exact copies of training documents. Instead, it functions as a statistical engine that ingests vast amounts of information, learns the underlying patterns, and creates completely new, probabilistic outputs based on those learned rules.Box 3: Yes A generative AI model produces output by predicting the next token based on patterns learned from the model’s training data.This is exactly how it works. Generative AI models (especially Large Language Models) generate content by continuously predicting what segment of data (a token) comes next based on the statistical patterns learned during their training.
https://learn.microsoft.com/en-us/azure/developer/ai/gen-ai-concepts-considerations-developers https://www.ibm.com/think/topics/generative-model
You need to build an AI solution that produces new product images based on written descriptions provided by users.Which AI workload should you use?
Answer(s): B
An AI workload that creates entirely new visual content based on written natural language descriptions is defined as image generation, which is a core capability of Generative AI. In the Microsoft Azure ecosystem, this task is primarily handled by generative models like DALL-E or GPT-image via Azure OpenAI Service and Azure AI Foundry.
https://learn.microsoft.com/en-us/azure/architecture/data-guide/ai-services/image-video-processing
Share your comments for Microsoft AI-901 exam with other users:
Finally got a change to write this exam and pass it! Valid and accurate!
Upload this exam please!
Thank you for providing these questions. It helped me a lot with passing my exam.
my first attempt
very explainable
i think answer of q 462 is variance analysis
hi i need see questions
best study material for exam
very interesting repository
american history 1
good level of questions
i need this dump kindly upload it
do we need c# coding to be az204 certified
excellent topics covered
are these really financial cloud questions and answers, seems these are basic admin question and answers
are these comments real
please upload the latest dumps
a company runs its workloads on premises. the company wants to forecast the cost of running a large application on aws. which aws service or tool can the company use to obtain this information? pricing calculator ... the aws pricing calculator is primarily used for estimating future costs
looks interesting
thanks! that’s amazing
the exam dumps are helping me get a solid foundation on the practical techniques and practices needed to be successful in the auditing world.
q 14 should be dmz sever1 and notepad.exe why does note pad have a 443 connection
question # 108, correct answers are business growth and risk reduction.
are these valid chfi questions
question: 162 should be dlp (b)
good exam questions
I have to say this is really close to real exam. Passed my exam with this.
good analytics question
this looks accurate
question 46, the answer should be data "virtualization" (not visualization).
its useful.
Pass this exam 3 days ago. The PDF version and the Xengine App is quite useful.
informative for me.
question 134s answer shoule be "dlp"
Keeping this site free takes real effort. We constantly battle automated scraping and unauthorized content copying. A quick account helps us protect the community and keep the site free.
To continue studying for your AI-901, please sign in or create a free account.