Microsoft · Associate
Microsoft Machine Learning Operations Engineer Associate (AI-300)
Run MLOps and GenAIOps on Azure Machine Learning and Microsoft Foundry.
Replaces DP-100. Microsoft retired DP-100 (Azure Data Scientist Associate) on June 1, 2026; AI-300 is the exam to take now.Source: learn.microsoft.com/en-us/credentials/support/retired-certification-exams and techcommunity.microsoft.com/blog/skills-hub-blog/the-ai-job-boom-is-here-are-you-ready-to-showcase-your-skills/4494128, verified 2026-10-11
AI-300 exam at a glance
| Exam code | AI-300 |
|---|---|
| Vendor | Microsoft |
| Level | Associate |
| Domains | 5 |
| Published practice questions | 251 |
| Our mock exam | 50 questions in 120 minutes |
| Our practice pass target | 70% |
| Checked against the exam guide | August 25, 2026 |
| Official exam page | Microsoft — AI-300 |
The mock exam length, time limit and pass target are our practice settings, not official exam scores or a guarantee of passing. Check the official AI-300 exam page for the current format, price and scoring before you book.
AI-300 exam domains
Weights show our practice allocation, built from the official exam guide; official blueprints may give ranges. Each domain links to its free sample questions (with explanations) and study guide preview once they are published.
Design and implement an MLOps infrastructure
19%Set up Azure Machine Learning workspaces, assets, and infrastructure as code.
- Workspaces, datastores, compute targets creation and management
- Identity and access management for workspaces
- Data assets, environments, components; registries for sharing
- GitHub integration and secure access
- IaC with Bicep and Azure CLI for ML resources
- GitHub Actions workflows for resource provisioning
- Network restriction of workspaces
- Git source control for ML projects
Free Design and implement an MLOps infrastructure practice questions →Design and implement an MLOps infrastructure study guide →
Implement machine learning model lifecycle and operations
29%Orchestrate training, registration, deployment, and monitoring of ML models.
- MLflow experiment tracking; automated ML; notebooks
- Hyperparameter tuning automation and distributed training
- Training pipelines and comparing model performance across jobs
- Model registration/versioning, feature retrieval specs, responsible AI evaluation
- Real-time vs batch endpoints with managed inference
- Progressive rollout and safe rollback
- Data drift detection and production performance monitoring
- Retraining triggers and alerts
Free Implement machine learning model lifecycle and operations practice questions →Implement machine learning model lifecycle and operations study guide →
Design and implement a GenAIOps infrastructure
24%Configure Microsoft Foundry environments and deploy foundation models for production.
- Foundry resources and project environments
- Managed identities and RBAC for GenAI workloads
- Network security and private networking configurations
- Bicep/Azure CLI deployment of Foundry infrastructure
- Serverless API endpoints vs managed compute for foundation models
- Model selection and versioning for production
- Provisioned throughput units for high-volume workloads
- Prompt versioning and management with Git
Free Design and implement a GenAIOps infrastructure practice questions →Design and implement a GenAIOps infrastructure study guide →
Implement generative AI quality assurance and observability
14%Evaluate and observe generative AI applications and agents.
- Test datasets and data mapping for evaluation
- AI quality metrics: groundedness, relevance, coherence, fluency
- Risk and safety evaluations for harmful content
- Automated evaluation workflows with built-in and custom evaluators
- Continuous monitoring in Foundry
- Latency, throughput, response-time monitoring
- Token consumption and cost tracking
- Logging, tracing, and production debugging
Free Implement generative AI quality assurance and observability practice questions →Implement generative AI quality assurance and observability study guide →
Optimize generative AI systems and model performance
14%Optimize RAG performance and implement advanced fine-tuning.
- Tuning similarity thresholds, chunk sizes, retrieval strategies
- Embedding model selection and fine-tuning for domains
- Hybrid search combining semantic and keyword retrieval
- RAG evaluation with relevance metrics and A/B testing
- Advanced fine-tuning methods
- Synthetic data creation and management for fine-tuning
- Fine-tuned model lifecycle from development to production
Free Optimize generative AI systems and model performance practice questions →Optimize generative AI systems and model performance study guide →
AI-300 FAQ
How is the Machine Learning Operations Engineer Associate (AI-300) practice assessment configured?
Our practice format targets 50 questions in 120 minutes. Availability depends on the published question bank and is shown on this course page. Official exam question counts and formats may differ.
What does the 70% practice target mean?
Microsoft technical exams use a scaled passing score of 700; this is not equivalent to 70% correct. DataCertPrep uses 70% as an internal practice target, not a prediction or guarantee of passing.
How should I assess my progress in Machine Learning Operations Engineer Associate (AI-300)?
Use practice accuracy to find weak topics. Repeating familiar questions can inflate your score. Check coverage of the current objectives, practise hands-on tasks, and assess yourself on unseen questions.
What is the best way to study Machine Learning Operations Engineer Associate (AI-300)?
Read the available guides, work through practice explanations, use flashcards and complete practical exercises. Take a timed assessment when available. Check the provider's current exam guide before booking your exam.
Should I use AI-300 exam dumps?
No. "Dumps" are questions copied or recalled from the live exam. Certification exams are covered by a candidate agreement that prohibits sharing or using them, and breaking it can lead to a revoked certification or a ban from future exams. Dumps are also often wrong, and they go out of date when the exam blueprint changes. Memorising answers does not build the skills the exam tests. Study the official Microsoft exam guide and practise with original questions that explain each answer, such as the free sample questions linked on this page.
What happened to DP-100?
Microsoft retired DP-100 (Azure Data Scientist Associate) on June 1, 2026, and AI-300 is the exam that replaces it. If you studied for DP-100, use the AI-300 domains below to see what to review.