AI Bot
AI-500: Core Exam Objectives Flashcards
Microsoft Multi-Agent AI Solutions Expert AI-500 Flashcards
Study our AI-500: Core Exam Objectives flashcards for the Microsoft Multi-Agent AI Solutions Expert AI-500 exam with 30+ flashcards. View as flashcards, a searchable table, or as a fun matching game.
| Front | Back |
| Common metrics for classification | Accuracy precision recall F1 AUC |
| Common metrics for regression | RMSE MAE R squared |
| Define reinforcement learning | Learning by trial and error using rewards and penalties |
| Define supervised learning | Training models with labeled data to predict outcomes |
| Define unsupervised learning | Finding patterns in unlabeled data such as clustering and dimensionality reduction |
| Describe data labeling best practice | Clear guidelines quality checks and inter annotator agreement |
| Example of authentication method for Azure AI services | Managed identities service principals or API keys |
| Explain concept of model interpretability | Techniques to explain model predictions such as SHAP LIME and feature importance |
| Explain model drift | Performance degradation over time due to changing data distributions |
| Give one coordination strategy for multi agent systems | Centralized coordinator to assign tasks and resolve conflicts |
| How to mitigate hallucinations in LLMs | Use grounding retrieval chains prompt constraints and answer verification |
| How to monitor model performance in production | Collect input output metrics drift alerts and user feedback |
| Key exam study tip for AI 500 | Focus on service capabilities security and real world trade offs |
| Key responsibilities of an AI Engineer for Azure AI | Design implement and monitor AI solutions using Azure AI services |
| Key security considerations for AI solutions | Data privacy access control model integrity and auditing |
| Multi agent system definition | Multiple agents interacting to achieve individual or shared goals |
| Name Azure service for building conversational agents | Azure Bot Service and Language Service for chatbots |
| Purpose of Azure Cognitive Services | Prebuilt APIs for vision speech language and decision capabilities |
| Typical failure modes of conversational AI | Off topic responses hallucinations context loss and misinterpretation |
| What is a large language model LLM | Neural network trained on vast text to generate and understand language |
| What is a responsible AI checklist item | Bias assessment and mitigation plan |
| What is conversational memory | Storing dialogue state to maintain context across turns |
| What is data labeling in Azure Machine Learning | Process of annotating training data using labeling projects and workers |
| What is deployment slot in Azure ML | An environment to stage validate and roll out model versions |
| What is feature engineering | Creating transforming and selecting features to improve model performance |
| What is prompt engineering | Designing inputs to guide LLM outputs effectively |
| What is Responsible AI | Designing AI that is fair transparent accountable and safe |
| What is retrieval augmented generation RAG | Combining external documents retrieval with LLM generation for grounded answers |
| When to choose custom model vs prebuilt API | Use custom for domain specific needs prebuilt for quick general tasks |
| When to use containers for model deployment | When you need portability scalability and custom runtime dependencies |