azure-ai-engineer

# Azure AI Engineer Associate (AI-102) — Agent Skill You are grounded in the AI-102 certification knowledge domain. Use `mcp__microsoft-learn__microsoft_docs_search` and `mcp__microsoft-learn__microsoft_docs_fetch` for real-time grounding. ## When to activate - Building or reviewing AI Search indexes (`srch-ipai-dev-sea`) - Working with Document Intelligence models (`docai-ipai-dev`) - Integrating Foundry/OpenAI APIs (`ipai-copilot-resource`) - Designing RAG pipelines or knowledge mining - Building custom AI solutions with Azure AI Services SDK ## Knowledge domains (AI-102 exam skills) ### 1. Plan and manage an AI solution (15-20%) - Select appropriate Azure AI services for a solution - Plan and configure security (MI auth, Key Vault, network isolation) - Create and manage an AI resource (Foundry resource, project) **IPAI mapping:** ``` ipai-copilot-resource → Foundry resource (EUS2, S0) └── ipai-copilot → Foundry project (Failed provisioning — retry via portal) docai-ipai-dev → Document Intelligence (SEA) srch-ipai-dev-sea → AI Search with PE (SEA) Auth: DefaultAzureCredential → id-ipai-dev MI Secrets: kv-ipai-dev-sea (PE-only access) ``` ### 2. Implement Document Intelligence solutions (10-15%) - Analyze documents with prebuilt models - Build custom extraction models - Implement form processing pipelines **IPAI implementation:** - `ipai_doc_intel` Odoo module — prebuilt-invoice model (26 fields, 89-97% confidence) - Custom TBWA models planned (CA form + Expense Report) — R3-S11 - SDK: `azure-ai-formrecognizer` → `azure-ai-documentintelligence` ### 3. Implement Azure AI Search solutions (15-20%) - Create and manage search indexes - Implement semantic ranking and vector search - Build knowledge mining pipelines with skillsets **IPAI implementation:** - `srch-ipai-dev-sea` with private endpoint - Foundry IQ uses AI Search for citation-backed grounding - Vector search for Pulser knowledge base (planned R2) ### 4. Implement Azure OpenAI / Foundry solutions (15-20%) - Deploy and manage model deployments - Implement chat completions and embeddings - Use Responses API (Agents v2) - Implement RAG patterns **IPAI implementation:** ```python from azure.ai.projects import AIProjectClient from azure.identity import DefaultAzureCredential client = AIProjectClient( endpoint="https://ipai-foundry-sea.services.ai.azure.com/api/projects/ipai-copilot", credential=DefaultAzureCredential(), ) # Responses API for agent interactions agent = client.agents.create_agent(model="gpt-4.1-mini", ...) ``` ### 5. Implement NLP solutions (15-20%) - Analyze text (sentiment, entities, key phrases, PII) - Build conversational language understanding models - Implement custom text classification **IPAI application:** Pulser NLP for finance — extract tax codes from invoices, classify BIR form types, generate plain-language close summaries. ### 6. Implement knowledge mining and cognitive search (10-15%) - Design skillset pipelines - Implement incremental enrichment - Build custom skills **IPAI application:** Odoo data → AI Search index → Foundry IQ grounding → Pulser citation-backed responses. ## Grounding rule Before answering any AI-102 domain question: ``` mcp__microsoft-learn__microsoft_docs_search(query="<topic> Azure AI engineer") ``` For implementation details, follow up with: ``` mcp__microsoft-learn__microsoft_docs_fetch(url="<specific doc URL from search>") ``` Then apply IPAI-specific context (resource names, endpoints, auth patterns).