About this guide
Python developers and Azure AI engineers preparing for AI-103, including those moving on from the retired AI-102.
Built for practical review: read each topic, test your understanding with original scenarios, and return to the official provider outline for any exam updates.
What’s inside
- Choosing Foundry models, tools, and deployment types
- Managed identity, private networking, quotas, and monitoring
- Responsible AI: content filters, prompt shields, and evaluators
- RAG with hybrid search and groundedness evaluation
- Agents, tool schemas, multi-agent patterns, and approvals
- Vision, speech, translation, and Content Understanding
- Thirty original practice questions and a six-week plan
What you’ll take away
- Design a keyless, private, and monitored Foundry solution.
- Build and evaluate a grounded RAG app and a safe agent.
- Choose the right Foundry Tool for vision, text, speech, and extraction tasks.
Frequently asked questions
Is AI-103 the replacement for AI-102?
AI-102 is retired in the PrepKloud catalogue, and AI-103 covers the current Microsoft Foundry apps-and-agents skills. Check Microsoft Learn for official transition details.
What background do I need?
Microsoft expects Python development experience and familiarity with generative AI and Azure services.
Does the guide follow the latest skills outline?
It follows the skills measured as of 16 April 2026, including all five skill areas and their weightings.