Dashboard | AI-103 — CertRunway

Completion0%0 of 250 cards studied

Know00% of Study Deck

Unsure00% of Study Deck

Missed00% of Study Deck

Recommended Study Strategy

Round 1

Go through all 250 Study Deck cards in order and grade each one Know, Unsure, or Missed. This sets your baseline.

Remaining: 250 cards

Round 2

Study only the cards you graded Unsure or Missed. A card you grade Know leaves the Review Deck the next time you start a session.

Active pool: 0 cards

Round 3

Shuffle the full Study Deck and grade from memory to test what you really know.

Pass mark: 225 of 250 cards (90%)

Match the endpoint to the client: a project endpoint (services.ai.azure.com/api/projects/...) goes with AIProjectClient, and the Azure OpenAI v1 endpoint (openai.azure.com/openai/v1/) goes with the OpenAI client. Prefer keyless Microsoft Entra ID over keys.
Know who does what in an agent: the model decides which tool to call, your app runs a function call and returns its output, and long jobs such as video generation and document analysis are asynchronous and polled.
Pick the simplest orchestration that works: one agent with tools first, sequential for fixed steps, concurrent for independent parallel work, handoff when the right specialist isn't known upfront, and group chat for maker-checker review.
For grounding, hybrid search with semantic ranking is the recommended retrieval, and an agent only searches and cites reliably when its instructions tell it to.
Expect renamed and retiring features: the Azure AI roles are now Foundry roles, visual workflows retire December 1, 2026, and several Azure Language features, including sentiment analysis and summarization, retire March 31, 2029.
Read the final sentence of a long scenario first to find the actual question.
Watch for qualifiers such as NOT, BEST, and MOST appropriate.
You have 120 minutes and need 700 to pass. Flag hard questions and come back; change an answer only with a clear reason.