Deep Dive: AI-102 Azure AI Engineer â Complete Knowledge Structure (Low â High)
Deep Dive: AI-102 Azure AI Engineer â Complete Knowledge Structure (Low â High)
Topic: AI-102: Designing and Implementing a Microsoft Azure AI Solution
Category: Artificial Intelligence / Cloud AI Services
Level: Beginner â Intermediate â Advanced (Progressive)
Last Updated: 2026-03-15
äžæç (Chinese Version)
1. æŠè¿° (Overview)
AI-102 æ¯ Microsoft Azure AI Engineer Associate 讀è¯ç宿¹è¯ŸçšïŒé¢ååžæåš Azure äžæå»º AI è§£å³æ¹æ¡ç蜯件åŒåè ãè¯Ÿçšæ¶µç 5 倧åŠä¹ è·¯åŸã40 䞪暡åïŒå æ¬ïŒçæåŒ AI åºçšåŒåãAI Agent æå»ºãèªç¶è¯èšå€çãè®¡ç®æºè§è§ã以åä¿¡æ¯æåã
æ¬æå° AI-102 çå šéšç¥è¯äœç³»ä» Level 0ïŒåºç¡ïŒå° Level 7ïŒé«çº§ïŒ éå±å±åŒïŒåž®å©äœ 建ç«å®æŽçç¥è¯å°åŸãæ¯äžå±éœå»ºç«åšåäžå±çåºç¡äžïŒç¡®ä¿äœ èœåŸªåºæžè¿å°ææ¡ææå ³é®æèœã
è¯Ÿçšæ»è§ â 5 倧åŠä¹ è·¯åŸ
| # | åŠä¹ è·¯åŸ | æš¡åæ° | æ žå¿äž»é¢ |
|---|---|---|---|
| 1 | Develop Generative AI Apps in Azure | 8 | çæåŒ AIãRAGãFine-tuningãPrompt Flow |
| 2 | Develop AI Agents on Azure | 9 | Agent ServiceãMCPãMulti-agentãA2A |
| 3 | Develop Natural Language Solutions | 10 | ææ¬åæãCLUãè¯é³ãç¿»è¯ |
| 4 | Develop Computer Vision Solutions | 8 | åŸååæãOCRã人èžãèªå®ä¹è§è§ |
| 5 | Develop AI Information Extraction Solutions | 5 | Document IntelligenceãAI SearchãContent Understanding |
2. Level 0 â åºç¡å¹³å°äžåå€ (Foundation & Setup)
ð¯ ç®æ ïŒçè§£ Azure AI çæïŒæå»ºåŒåç¯å¢
2.1 Azure AI æå¡å šæ¯åŸ
Azure AI æå¡äœç³»çæ žå¿ç»æïŒ
âââââââââââââââââââââââââââââââââââââââââââââââââââââââ
â Microsoft Foundry â
â (ç»äžç AI åŒåå¹³å°ïŒå Azure AI Studio) â
â â
â ââââââââââââ ââââââââââââ ââââââââââââ â
â â Model â â Prompt â â Agent â â
â â Catalog â â Flow â â Service â â
â ââââââââââââ ââââââââââââ ââââââââââââ â
â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â Azure AI Services â â
â â ââââââââââ ââââââââââ ââââââââââ âââââââââââ â
â â âLanguage â âVision â âSpeech â âDocumentââ â
â â âService â âService â âService â âIntelli.ââ â
â â ââââââââââ ââââââââââ ââââââââââ âââââââââââ â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â Azure OpenAI Service â â
â â GPT-4o | GPT-4 | DALL-E | Whisper â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â Azure AI Search â â
â â çŽ¢åŒ | æèœé | åéæçŽ¢ | è¯ä¹æåº â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
âââââââââââââââââââââââââââââââââââââââââââââââââââââââ
2.2 å ³é®åºç¡æŠå¿µ
| æŠå¿µ | 诎æ |
|---|---|
| Microsoft Foundry | ç»äž AI åŒåå¹³å°ïŒå Azure AI StudioïŒïŒç®¡çæš¡åãæ°æ®ãéšçœ² |
| Azure AI Services | é¢æå»ºè®€ç¥æå¡çéåïŒè¯èšãè§è§ãè¯é³ãå³çïŒ |
| Azure OpenAI Service | éè¿ Azure è®¿é® OpenAI æš¡åïŒGPTãDALL-EãWhisper çïŒ |
| Resource & Endpoint | æ¯äžª AI æå¡éœéèŠåå»ºèµæºïŒéè¿ Endpoint + Key æ Entra ID è®¿é® |
| REST API & SDK | 䞀ç§è°çšæ¹åŒïŒçŽæ¥ HTTP è°çš REST APIïŒæäœ¿çš Python/C# SDK |
2.3 åŒåç¯å¢åå€
åçœ®èŠæ±ïŒ
âââ Azure 订é
ïŒå¯çšå
莹è¯çšïŒ
âââ Python 3.8+ æ C#/.NET
âââ Visual Studio Code
â âââ Azure AI Foundry æ©å±
â âââ Python / C# æ©å±
âââ Azure CLI
âââ Azure AI Foundry SDK (pip install azure-ai-projects)
2.4 讀è¯äžææ
- API Key æ¹åŒïŒç®åäœäžæšèçšäºç产
- Microsoft Entra IDïŒå AADïŒïŒæšèçç产æ¹åŒïŒäœ¿çš
DefaultAzureCredential - RBAC è§è²ïŒ
Cognitive Services UserãCognitive Services Contributor
3. Level 1 â é¢æå»º AI æå¡ïŒåŒç®±å³çšïŒ
ð¯ ç®æ ïŒåŠäŒçŽæ¥è°çš Azure é¢æå»º AI èœåïŒæ éè®ç»æš¡å
3.1 ææ¬åæ (Text Analytics)
æå¡ïŒAzure Language ServiceïŒFoundry Tools äžç Language åèœïŒ
| åèœ | 诎æ | å žåçšé |
|---|---|---|
| æ æåæ (Sentiment Analysis) | å€æææ¬æ£é¢/èŽé¢/äžæ§æ 绪 | 客æ·åéŠåæã瀟亀åªäœçæ§ |
| å ³é®çè¯æå (Key Phrase Extraction) | æåææ¬äžçå ³é®è¯åçè¯ | ææ¡£æèŠãæ çŸçæ |
| åœåå®äœè¯å« (NER) | è¯å«äººåãå°ç¹ãç»ç»ãæ¥æç | ä¿¡æ¯æåãæ°æ®ç»æå |
| è¯è𿣿µ (Language Detection) | å€æææ¬äœ¿çšçè¯èš | å€è¯èšç³»ç»è·¯ç± |
| å®äœéŸæ¥ (Entity Linking) | å°å®äœéŸæ¥å°ç»ŽåºçŸç§æ¡ç® | ç¥è¯åŸè°±ãæ¶æ§ |
| PII æ£æµ | è¯å«äžªäººææä¿¡æ¯ | æ°æ®è±æãåè§ |
è°çšæš¡åŒïŒ
客æ·ç«¯ â REST API / SDK â Language Endpoint â è¿å JSON ç»æ
3.2 ç¿»è¯æå¡ (Translator Service)
- ææ¬ç¿»è¯ïŒæ¯æ 100+ è¯èšç宿¶ç¿»è¯
- ææ¡£ç¿»è¯ïŒæ¹éç¿»è¯æŽäžªææ¡£ïŒä¿çæ ŒåŒ
- èªå®ä¹ç¿»è¯åšïŒäœ¿çšèªå·±çæ¯è¯è¡šè®ç»å®å¶ç¿»è¯æš¡å
- é³è¯ (Transliteration)ïŒåšäžåæåç³»ç»éŽèœ¬æ¢ïŒåŠæ¥æååâçœé©¬åïŒ
3.3 è¯é³æå¡ (Speech Services)
| åèœ | æ¹å | 诎æ |
|---|---|---|
| Speech-to-Text (STT) | è¯é³ â ææ¬ | 宿¶/æ¹éè¯é³è¯å« |
| Text-to-Speech (TTS) | ææ¬ â è¯é³ | èªç¶è¯é³åæïŒæ¯æèªå®ä¹è¯é³ |
| Speech Translation | è¯é³ â ç¿»è¯ææ¬ | 宿¶è¯é³ç¿»è¯ |
| Speaker Recognition | è¯é³ â 身仜 | 诎è¯äººéªè¯åè¯å« |
å ³é®æŠå¿µïŒ
- Speech ConfigïŒé 眮订é å¯é¥ååºå
- Audio ConfigïŒé 眮é³é¢èŸå ¥/èŸåºïŒéºŠå é£ãæä»¶ãæµïŒ
- SSMLïŒSpeech Synthesis Markup LanguageïŒç²Ÿç»æ§å¶è¯é³åæ
3.4 åŸååæ (Image Analysis)
æå¡ïŒAzure Vision Service
| åèœ | 诎æ |
|---|---|
| åŸåæè¿° | èªåšçæåŸåçèªç¶è¯èšæè¿° |
| æ çŸæå | è¯å«åŸåäžç对象ãåºæ¯ãåšäœ |
| å¯¹è±¡æ£æµ | å®äœåŸåäžå¯¹è±¡çäœçœ®ïŒèŸ¹çæ¡ïŒ |
| æºèœè£åª | åºäºå ³æ³šåºåèªåšè£åª |
| äººèžæ£æµ | æ£æµäººèžäœçœ®å屿§ |
| OCRïŒå åŠå笊è¯å«ïŒ | ä»åŸåäžæåææ¬ïŒè§ 3.5ïŒ |
3.5 OCR â 读ååŸåäžçææ¬
äž€ç§ APIïŒ
- Image Analysis Read APIïŒç®ååºæ¯ïŒåæ¥ïŒ
- Document IntelligenceïŒå€æææ¡£ïŒåŒæ¥ïŒè§ Level 3ïŒ
æµçšïŒ
åŸåèŸå
¥ â é¢å€ç â ææ¬æ£æµ â æåè¯å« â è¿åç»æåææ¬
(è¡ãåè¯ãèŸ¹çæ¡)
æ¯æïŒæå°ææ¬ãæåææ¬ãå€è¯èšæ··å
3.6 äººèžæ£æµäžè¯å« (Face Detection & Recognition)
äžå±èœåïŒ
- Face DetectionïŒæ£æµäººèžäœçœ®å屿§ïŒé审æ¹ïŒ
- Face VerificationïŒ1:1 æ¯å¯¹ïŒè¿äž€åŒ ç §çæ¯åäžäžªäººåïŒïŒ
- Face IdentificationïŒ1:N è¯å«ïŒè¿äžªäººæ¯è°ïŒïŒ
â ïž èŽèޣ任 AI 泚æïŒäººèžè¯å«åèœåéè®¿é®æ¿ç纊æïŒéèŠç³è¯·å®¡æ¹
3.7 è§é¢åæ (Video Indexer)
Azure Video Indexer ä»è§é¢äžæåå€ç»ŽæŽå¯ïŒ
- 人èžè¯å«äžè·èžª
- OCRïŒè§é¢äžçæåïŒ
- è¯é³èœ¬ææ¬
- äž»é¢/å ³é®è¯æå
- æ æåæ
- åºæ¯åå²
- å å®¹å®¡æ ž
3.8 é¢æå»º Document Intelligence æš¡å
| æš¡å | æåå 容 |
|---|---|
| å祚暡å | å祚å·ãæ¥æãéé¢ãäŸåºåä¿¡æ¯ |
| æ¶æ®æš¡å | åå®¶åãäº€ææ¥æãæç»ãæ»é¢ |
| èº«ä»œè¯æš¡å | å§åãåºçæ¥æãè¯ä»¶å· |
| åçæš¡å | å§åãèäœãå ¬åžãèç³»æ¹åŒ |
| çšå¡ææ¡£ | W-2ã1098 ççŸåœçšå¡è¡šæ Œ |
4. Level 2 â èªå®ä¹ AI æš¡åïŒè®ç»äœ èªå·±çæš¡åïŒ
ð¯ ç®æ ïŒåšé¢æå»ºèœåäžè¶³æ¶ïŒè®ç»éåäœ äžå¡åºæ¯çèªå®ä¹æš¡å
4.1 对è¯è¯èšçè§£ (Conversational Language Understanding â CLU)
æ žå¿æŠå¿µïŒ
| æŠå¿µ | 诎æ | ç€ºäŸ |
|---|---|---|
| UtteranceïŒè¯è¯ïŒ | çšæ·è¯Žçäžå¥è¯ | âåž®æé¢è®¢æå€©å»åäº¬çæºç¥šâ |
| IntentïŒæåŸïŒ | çšæ·æ³åä»ä¹ | BookFlight |
| EntityïŒå®äœïŒ | è¯è¯äžçå ³é®åæ° | æ¥æ=æå€©, ç®çå°=å京 |
è®ç»æµçšïŒ
1. å®ä¹ Intent å Entity
2. æ æ³šè®ç»æ°æ®ïŒUtterance â Intent + EntityïŒ
3. è®ç»æš¡å
4. è¯äŒ°ïŒPrecision / Recall / F1ïŒ
5. éšçœ²æš¡å
6. åºçšéæ
4.2 é®çæå¡ (Question Answering)
åºäºç¥è¯åºçé®çç³»ç»ïŒ
- æ°æ®æºïŒFAQ çœé¡µãWord/PDF ææ¡£ãæåšæ·»å QA 对
- å€èœ®å¯¹è¯ïŒæ¯æ follow-up prompts å®ç°åŒå¯ŒåŒå¯¹è¯
- 粟确åçïŒä»æ®µèœäžç²Ÿç¡®æåçæ¡çæ®µ
- åä¹è¯ïŒé 眮 alterations æåå¹é ç
4.3 èªå®ä¹ææ¬åç±» (Custom Text Classification)
| ç±»å | 诎æ |
|---|---|
| åæ çŸåç±» | æ¯äžªææ¡£åªå±äºäžäžªç±»å« |
| 倿 çŸåç±» | æ¯äžªææ¡£å¯å±äºå€äžªç±»å« |
è®ç»èŠæ±ïŒ
- è³å° 10 äžªææ¡£/ç±»å«
- æšè 200+ äžªææ¡£/ç±»å«
- éèŠæ æ³šæ°æ®
4.4 èªå®ä¹åœåå®äœè¯å« (Custom NER)
ä»éç»æåææ¬äžæåèªå®ä¹å®äœç±»åïŒ
- å®ä¹äœ çå®äœç±»åïŒåŠ â产ååâãâååçŒå·âïŒ
- æ æ³šè®ç»æ°æ®
- è®ç» â è¯äŒ° â éšçœ²
- è¯äŒ°ææ ïŒPrecisionãRecallãF1-Score
4.5 èªå®ä¹è§è§ â åŸååç±» (Custom Vision - Classification)
è®ç»éåŸå â äžäŒ å° Custom Vision â éæ©å类类å â è®ç» â è¯äŒ° â ååž
âââ å€ç±»å«ïŒæ¯åŸäžäžªæ çŸïŒ
âââ 倿 çŸïŒæ¯åŸå€äžªæ çŸïŒ
æäœ³å®è·µïŒ
- æ¯ç±»è³å° 15 åŒ åŸ
- å å«äžåè§åºŠãå ç §ãèæ¯
- äœ¿çš Smart Labeler åèªåšæ æ³š
4.6 èªå®ä¹è§è§ â å¯¹è±¡æ£æµ (Custom Vision - Object Detection)
äžåç±»çåºå«ïŒäžä» è¯å«æ¯ä»ä¹ïŒè¿å®äœåšåªéïŒèŸ¹çæ¡ïŒ
- éèŠæ æ³šèŸ¹çæ¡ïŒBounding BoxïŒ
- éçšåºæ¯ïŒå·¥äžèŽšæ£ã莧æ¶ååè¯å«ãå®é²çæ§
4.7 èªå®ä¹ Document Intelligence æš¡å
åœé¢æå»ºæš¡åæ æ³æ»¡è¶³éæ±æ¶ïŒ
- èªå®ä¹æš¡æ¿æš¡åïŒéçšäºåºå®æ ŒåŒè¡šå
- èªå®ä¹ç¥ç»æš¡åïŒéçšäºåç»æååéç»æåææ¡£
- ç»åæš¡åïŒç»åå€äžªæš¡åå€çäžå衚åç±»å
5. Level 3 â çæåŒ AI åºç¡ (Generative AI Fundamentals)
ð¯ ç®æ ïŒææ¡ LLM æš¡åéæ©ãéšçœ²ååºæ¬åºçšåŒå
5.1 æš¡åç®åœäžéšçœ² (Model Catalog & Deployment)
Microsoft Foundry ç Model Catalog æäŸå€ç§æš¡åïŒ
| æš¡åæ¥æº | ç€ºäŸ | éšçœ²æ¹åŒ |
|---|---|---|
| Azure OpenAI | GPT-4o, GPT-4, GPT-3.5-turbo | æ å/å šå±éšçœ² |
| Meta | Llama 3 | Serverless API / Managed Compute |
| Mistral | Mistral Large, Mixtral | Serverless API |
| Microsoft | Phi-3, Phi-4 | Serverless API / Managed Compute |
| Cohere | Command R+ | Serverless API |
éšçœ²ç±»åïŒ
- Serverless API (MaaS)ïŒæ Token ä»èŽ¹ïŒæ é管çåºç¡è®Ÿæœ
- Managed Compute (MaaP)ïŒç¬å 计ç®èµæºïŒéåé«åå
- æ åéšçœ²ïŒAzure OpenAI æš¡åçé»è®€æ¹åŒ
5.2 Microsoft Foundry SDK
from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential
# å建项ç®å®¢æ·ç«¯
client = AIProjectClient(
credential=DefaultAzureCredential(),
endpoint="https://<your-foundry>.services.ai.azure.com"
)
# è°çšè倩宿
response = client.inference.get_chat_completions(
model="gpt-4o",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is Azure AI?"}
]
)
æ žå¿ç»ä»¶ïŒ
AIProjectClientïŒé¡¹ç®çº§å®¢æ·ç«¯ïŒç®¡çèµæºåè¿æ¥ChatCompletionsClientïŒè倩宿è°çšEmbeddingsClientïŒææ¬åµå ¥çæ
5.3 Prompt Flow
Prompt Flow æ¯äžäžªå¯è§åç LLM åºçšçŒæå·¥å ·ïŒ
èŸå
¥ â [LLM èç¹] â [Python èç¹] â [LLM èç¹] â èŸåº
â â â
æç€ºæš¡æ¿ æ°æ®å€ç/API æ»ç»/æ ŒåŒå
æ žå¿æŠå¿µïŒ
- FlowïŒæµïŒïŒDAGïŒæåæ ç¯åŸïŒåœ¢åŒçå·¥äœæµ
- NodeïŒèç¹ïŒïŒLLM è°çšãPython 代ç ãå·¥å ·è°çš
- ConnectionïŒäžå€éšæå¡çè¿æ¥é 眮
- VariantïŒåäžèç¹çäžåæç€ºçæ¬ïŒçšäº A/B æµè¯
äžç§ Flow ç±»åïŒ
- Standard FlowïŒéçš LLM åºçšæµ
- Chat FlowïŒå¯¹è¯åŒåºçšïŒå 眮è倩åå²ç®¡ç
- Evaluation FlowïŒè¯äŒ°å ¶ä»æµç莚é
5.4 è§è§å¢åŒºççæåŒ AI åºçš
䜿çšå€æš¡ææš¡åïŒåŠ GPT-4oïŒå€çåŸåïŒ
response = client.chat.completions.create(
model="gpt-4o",
messages=[{
"role": "user",
"content": [
{"type": "text", "text": "Describe this image"},
{"type": "image_url", "image_url": {"url": image_url}}
]
}]
)
åºçšåºæ¯ïŒåŸåæè¿°ãè§è§é®çãåŸè¡šè§£è¯»ãææ¡£çè§£
5.5 è¯é³å¢åŒºççæåŒ AI åºçš
ç»å Speech Service å LLMïŒ
çšæ·è¯é³ â STT â ææ¬ â LLM â ååºææ¬ â TTS â è¯é³èŸåº
- 宿¶è¯é³å¯¹è¯ AI
- å€è¯èšè¯é³å©æ
- äŒè®®æèŠååæ
5.6 åŸåçæ (Image Generation)
äœ¿çš DALL-E æš¡åçæåŸåïŒ
- ææ¬å°åŸåïŒåºäºèªç¶è¯èšæè¿°çæåŸå
- åŸåçŒèŸïŒä¿®æ¹ç°æåŸåçç¹å®åºå
- åŸååäœïŒåºäºåèåŸåçæåäœ
6. Level 4 â é«çº§çæåŒ AI (Advanced Generative AI)
ð¯ ç®æ ïŒææ¡ RAGãFine-tuning åäŒäžçº§ç¥è¯æ£çŽ¢
6.1 RAG â æ£çŽ¢å¢åŒºçæ (Retrieval Augmented Generation)
䞺ä»ä¹éèŠ RAGïŒ
- LLM çè®ç»æ°æ®ææªæ¢æ¥æ
- LLM äžäºè§£äœ çç§ææ°æ®
- çŽæ¥ææææ°æ®å¡è¿ Prompt äŒè¶ åº Token éå¶
RAG æ¶æïŒ
ââââââââââââââââ
â äœ çæ°æ®æº â
â (ææ¡£/æ°æ®åº) â
ââââââââ¬ââââââââ
â 玢åŒé¶æ®µ
âŒ
ââââââââââââââââ
â Azure AI â
â Search Index â
â (åé+å
³é®è¯) â
ââââââââ¬ââââââââ
â æ¥è¯¢é¶æ®µ
çšæ·æé® âââââââââ â
⌠âŒ
ââââââââââââââââââââââââ
â æ£çŽ¢çžå
³ææ¡£ç段 â
ââââââââââââ¬ââââââââââââ
â
âŒ
ââââââââââââââââââââââââ
â ç»å Prompt: â
â System + Context â
â + User Question â
ââââââââââââ¬ââââââââââââ
â
âŒ
ââââââââââââââââââââââââ
â LLM çæåç â
â (åºäºæ£çŽ¢å°çäžäžæ) â
ââââââââââââââââââââââââ
å ³é®æ¥éª€ïŒ
- æ°æ®åå€ïŒææ¡£åçïŒChunkingïŒãåµå ¥çæ
- 玢åŒå建ïŒAzure AI Search å建åé玢åŒ
- æ£çŽ¢é çœ®ïŒæ··åæçŽ¢ïŒå ³é®è¯ + åé + è¯ä¹æåºïŒ
- Prompt å·¥çšïŒè®Ÿè®¡ System Prompt åŒå¯Œæš¡åäœ¿çšæ£çŽ¢äžäžæ
- ååºçæïŒLLM åºäºæ£çŽ¢ç»æçæææ ¹æ®çåç
Azure å®ç°ïŒ
- äœ¿çš Microsoft Foundry ç âAdd your dataâ åèœ
- æéè¿ SDK çŒçšå®ç°å®æŽ RAG 管é
6.2 Fine-tuning â æš¡å埮è°
äœæ¶éæ© Fine-tuning èé RAGïŒ
| åºæ¯ | RAG | Fine-tuning |
|---|---|---|
| éèŠææ°ç¥è¯ | â éŠé | â äžéå |
| æ¹åæš¡åè¡äžº/飿 Œ | â æé | â éŠé |
| åŠä¹ ç¹å®æ ŒåŒ/æš¡åŒ | â | â |
| åå° Token æ¶è | â | â |
| å¿«é宿œ | â | â éèŠæ¶éŽ |
Fine-tuning æµçšïŒ
1. åå€è®ç»æ°æ®ïŒJSONL æ ŒåŒçå¯¹è¯æ ·æ¬ïŒ
2. äžäŒ æ°æ®å° Foundry
3. éæ©åºç¡æš¡å
4. é
眮è®ç»åæ°ïŒepochs, learning rate, batch sizeïŒ
5. å¯åšè®ç»
6. è¯äŒ°åŸ®è°æš¡å
7. éšçœ²åŸ®è°æš¡å
6.3 Azure AI Search â ç¥è¯ææ
äžå€§èœåïŒ
| èœå | 诎æ |
|---|---|
| æ°æ®æå | ä» Blob StorageãSQLãCosmos DB çæåæ°æ® |
| AI å¢åŒº | äœ¿çš SkillsetïŒOCRãå®äœè¯å«ãç¿»è¯çïŒäž°å¯æ°æ® |
| æºèœæçŽ¢ | å šææçŽ¢ + åéæçŽ¢ + è¯ä¹æåº |
SkillsetïŒæèœéïŒâ å 眮æèœïŒ
- OCRãåŸååæãå ³é®çè¯æå
- å®äœè¯å«ãè¯è𿣿µãç¿»è¯
- ææ¡£æåãææ¬åå¹¶
- èªå®ä¹æèœïŒè°çšäœ èªå·±ç API
玢åŒå¢åŒºç®¡éïŒ
æ°æ®æº â Indexer â Skillset â å¢åŒºææ¡£ â 玢åŒ
â
âââââââââŽââââââââ
â Built-in â
â Skills â
â âââââââââââ â
â âOCR â â
â âNER â â
â âKeyPhraseâ â
â âTranslate â â
â âââââââââââ â
âââââââââââââââââ
6.4 Azure Content UnderstandingïŒå€æš¡æå 容åæïŒ
æ°äžä»£ä¿¡æ¯æåæå¡ïŒç»äžå€çå€ç§å 容类åïŒ
- ææ¡£ïŒæåç»æåä¿¡æ¯
- åŸåïŒåæè§è§å 容
- é³é¢ïŒèœ¬åœååæ
- è§é¢ïŒå€æš¡æç»Œååæ
äž Document Intelligence çå ³ç³»ïŒ
- Content Understanding æ¯æŽæ°çãæŽç»äžçæå¡
- Document Intelligence ä»å𿝿ïŒäžæ³šäºè¡šååææ¡£
7. Level 5 â AI Agent åŒå (AI Agent Development)
ð¯ ç®æ ïŒæå»ºèœèªäž»æ§è¡ä»»å¡çæºèœä»£ç
7.1 AI Agent æ žå¿æŠå¿µ
ä»ä¹æ¯ AI AgentïŒ
äŒ ç» LLM App: çšæ·æé® â LLM åç â ç»æ
AI Agent: çšæ·ç»ç®æ â Agent è§å â è°çšå·¥å
· â è§å¯ç»æ
â 忬¡è§å â 忬¡è°çš â ... â å®æç®æ
Agent = LLM + Tools + Memory + Planning
| ç»ä»¶ | 诎æ |
|---|---|
| LLMïŒå€§èïŒ | çè§£æä»€ãæšçãå³ç |
| ToolsïŒå·¥å ·ïŒ | æçŽ¢ãä»£ç æ§è¡ãAPI è°çšãæä»¶æäœ |
| MemoryïŒè®°å¿ïŒ | çæïŒå¯¹è¯äžäžæïŒåé¿æïŒæä¹ åååšïŒ |
| PlanningïŒè§åïŒ | å°å€æä»»å¡å解䞺æ¥éª€ |
7.2 Microsoft Foundry Agent Service
éè¿ Azure Portal æ VS Code å建å管ç AgentïŒ
from azure.ai.projects import AIProjectClient
from azure.ai.agents import Agent
# å建 Agent
agent = client.agents.create_agent(
model="gpt-4o",
name="my-assistant",
instructions="You are a helpful data analyst.",
tools=[
{"type": "code_interpreter"},
{"type": "file_search"}
]
)
å çœ®å·¥å ·ïŒ
- Code InterpreterïŒè¿è¡ Python 代ç ãæ°æ®åæãçæåŸè¡š
- File SearchïŒæçŽ¢äžäŒ çæä»¶å 容
- Bing GroundingïŒäœ¿çš Bing æçŽ¢è·å宿¶ä¿¡æ¯
- Azure AI SearchïŒæçŽ¢äŒäžç¥è¯åº
7.3 èªå®ä¹å·¥å ·éæ (Custom Tools)
åœå çœ®å·¥å ·äžå€æ¶ïŒå®ä¹äœ èªå·±ç FunctionïŒ
# å®ä¹å·¥å
· schema
tools = [{
"type": "function",
"function": {
"name": "get_weather",
"description": "Get current weather for a city",
"parameters": {
"type": "object",
"properties": {
"city": {"type": "string"}
},
"required": ["city"]
}
}
}]
# Agent äŒåšéèŠæ¶è°çšæ€å·¥å
·
# äœ èŽèŽ£å®ç°å®é
çåœæ°é»èŸå¹¶è¿åç»æ
7.4 MCP å·¥å ·éæ (Model Context Protocol)
MCP æ¯äžäžªåŒæŸæ ååè®®ïŒè®© Agent èœåšæåç°åè°çšå€éšå·¥å ·ïŒ
Agent ââ MCP Client ââ MCP Server ââ å€éšå·¥å
·/æå¡
- äŒå¿ïŒæ ååçå·¥å ·æ³šååè°çšåè®®
- åºæ¯ïŒè¿æ¥ç¬¬äžæ¹æå¡ãäŒäžå éšç³»ç»
7.5 Foundry IQ â ç¥è¯å¢åŒº Agent
Foundry IQ æäŸå ±äº«ç¥è¯å¹³å°ïŒå€äžª Agent å¯ä»¥è®¿é®ïŒ
âââââââââââââââââââââââââââââââââââ
â Foundry IQ â
â (äŒäžç¥è¯å
±äº«å¹³å°) â
â â
â ââââââââââââ ââââââââââââ â
â â Data â â RAG â â
â â Assets â â Pipeline â â
â ââââââââââââ ââââââââââââ â
ââââââââââââ¬âââââââââââââââââââââââ
â
ââââââââŒâââââââ
⌠⌠âŒ
Agent A Agent B Agent C
- æ°æ®äŒåæåæ£çŽ¢èŽšé
- é 眮 Agent æä»€ç¡®ä¿åŒçšæºçäžèŽæ§
- æ¯æå€ Agent å ±äº«åäžç¥è¯åº
7.6 äž Microsoft 365 éæ
å° Agent ååžå°ïŒ
- Microsoft TeamsïŒå¢éåäœäžçŽæ¥äœ¿çš
- Microsoft 365 CopilotïŒéæå° Copilot çæ
- Work IQïŒè®¿é® M365 äžçå·¥äœæ°æ®ïŒé®ä»¶ãæ¥åãææ¡£ïŒ
8. Level 6 â å€ Agent äžé«çº§çŒæ (Multi-Agent & Advanced Orchestration)
ð¯ ç®æ ïŒæå»ºå€ Agent åäœç³»ç»åäŒäžçº§ AI å·¥äœæµ
8.1 Microsoft Agent Framework (Semantic Kernel)
Semantic Kernel æ¯ Microsoft ç AI çŒæ SDKïŒ
import semantic_kernel as sk
kernel = sk.Kernel()
# æ·»å AI æå¡
kernel.add_service(AzureChatCompletion(
deployment_name="gpt-4o",
endpoint="https://...",
api_key="..."
))
# æ·»å æä»¶ïŒå·¥å
·ïŒ
kernel.add_plugin(TimePlugin(), "time")
kernel.add_plugin(MathPlugin(), "math")
# å建 Agent
agent = ChatCompletionAgent(
kernel=kernel,
name="analyst",
instructions="You are a data analyst..."
)
æ žå¿æŠå¿µïŒ
- KernelïŒAI çŒæçæ žå¿åŒæ
- PluginïŒå¯éçšçåèœæš¡å
- PlannerïŒèªåšå°ç®æ å解䞺æ¥éª€
- MemoryïŒåéååšåè¯ä¹è®°å¿
8.2 å€ Agent çŒæ (Multi-Agent Orchestration)
Agent Chat æš¡åŒïŒ
âââââââââââââââââââââââââââââââââââââââââââââââ
â Agent Group Chat â
â â
â ââââââââââââ ââââââââââââ ââââââââââââ â
â â Research â â Analyst â â Writer â â
â â Agent ââ â Agent ââ â Agent â â
â ââââââââââââ ââââââââââââ ââââââââââââ â
â â â â â
â æçŽ¢ä¿¡æ¯ åææ°æ® æ°åæ¥å â
â â
â åè°çç¥ïŒ â
â - SequentialïŒé¡ºåºïŒ â
â - Round-RobinïŒèœ®è¯¢ïŒ â
â - SelectionïŒæºèœéæ©ïŒ â
â - TerminationïŒå®ææ¡ä»¶ïŒ â
âââââââââââââââââââââââââââââââââââââââââââââââ
çŒæçç¥ïŒ | çç¥ | 诎æ | éçšåºæ¯ | |ââ|ââ|âââ| | Sequential | æåºå®é¡ºåºèœ®æµ | ç®åæµæ°Žçº¿ | | Round-Robin | 埪ç¯èœ®æµ | å¹³çåäž | | Selection Strategy | ç± LLM å³å®äžäžäžªåèšè | çµæŽ»åäœ | | Termination Strategy | å®ä¹å®ææ¡ä»¶ | èªåšåæ¢ |
8.3 A2A åè®® (Agent-to-Agent)
A2A æ¯äžäžªè·šå¹³å°ç Agent éŽéä¿¡åè®®ïŒ
Agent A ââ A2A Protocol ââ Agent B
(æ¬å°) (è¿çš/äžåå¹³å°)
æ žå¿åèœïŒ
- Agent DiscoveryïŒåç°è¿çš Agent çèœå
- Direct CommunicationïŒAgent éŽçŽæ¥éä¿¡
- Task CoordinationïŒåè°è·š Agent ç任塿§è¡
8.4 Agent 驱åšçå·¥äœæµ (Agent-Driven Workflows)
äœ¿çš Microsoft Foundry æå»ºæºèœå·¥äœæµïŒ
- çŒæå€äžª Agent åç»ä»¶
- æ¡ä»¶åæ¯å埪ç¯
- é误å€çåéè¯
- äººå·¥å®¡æ žèç¹ïŒHuman-in-the-loopïŒ
9. Level 7 â èŽèޣ任 AI äžç产å (Responsible AI & Production)
ð¯ ç®æ ïŒç¡®ä¿ AI è§£å³æ¹æ¡å®å šãå¯é ãèŽèޣ任
9.1 èŽèޣ任 AI åå (Responsible AI Principles)
Microsoft ç 6 倧 AI ååïŒ
| åå | 诎æ |
|---|---|
| å ¬å¹³æ§ (Fairness) | AI ç³»ç»åºå ¬å¹³å¯¹åŸ ææäºº |
| å¯é æ§äžå®å šæ§ (Reliability & Safety) | AI åºå¯é è¿è¡ïŒäžé æäŒ€å®³ |
| éç§äžå®å š (Privacy & Security) | ä¿æ€çšæ·æ°æ®åéç§ |
| å å®¹æ§ (Inclusiveness) | AI åºèµèœæ¯äžªäºº |
| éææ§ (Transparency) | AI å³çåºå¯çè§£ |
| é®èŽ£å¶ (Accountability) | 人类åºå¯¹ AI ç³»ç»èŽèŽ£ |
9.2 çæåŒ AI çå 容å®å š
Azure AI Content SafetyïŒ
çšæ·èŸå
¥ â èŸå
¥è¿æ»€åš â LLM â èŸåºè¿æ»€åš â è¿åç»çšæ·
â â
âââââââŽââââââ âââââââŽââââââ
â æ£æµ: â â æ£æµ: â
â - 仿š â â - 仿š â
â - æŽå â â - æŽå â
â - æ§å
容 â â - æ§å
容 â
â - èªæ® â â - èªæ® â
â - Jailbreakâ â - éè¯¯ä¿¡æ¯ â
âââââââââââââ âââââââââââââ
å䞪䞥é级å«ïŒSafe â Low â Medium â High
9.3 è¯äŒ°çæåŒ AI åºçš
å 眮è¯äŒ°ææ ïŒ
| ææ | è¯äŒ°å 容 |
|---|---|
| GroundednessïŒææ®æ§ïŒ | åçæ¯åŠåºäºæäŸçäžäžæ |
| RelevanceïŒçžå ³æ§ïŒ | åçæ¯åŠäžé®é¢çžå ³ |
| CoherenceïŒè¿èޝæ§ïŒ | åçæ¯åŠé»èŸè¿èޝ |
| FluencyïŒæµç æ§ïŒ | è¯èšæ¯åŠèªç¶æµç |
| SimilarityïŒçžäŒŒæ§ïŒ | äžåèçæ¡ççžäŒŒåºŠ |
| F1 Score | äžåèçæ¡çè¯æ±éå |
è¯äŒ°æ¹åŒïŒ
- æåšè¯äŒ°ïŒäººå·¥æå
- AI èŸ å©è¯äŒ°ïŒäœ¿çš LLM äœäžºè¯å€è
- èªåšåææ ïŒäœ¿çš Prompt Flow çè¯äŒ°æµ
10. ç¥è¯å°åŸæ»è§ (Complete Knowledge Map)
Level 7: èŽèޣ任 AI äžç产å
â èŽèޣ任 AI åå â å
容å®å
š â è¯äŒ°ææ
â
Level 6: å€ Agent äžé«çº§çŒæ
â Semantic Kernel â å€ Agent Chat â A2A åè®® â Agent å·¥äœæµ
â
Level 5: AI Agent åŒå
â Foundry Agent Service â èªå®ä¹å·¥å
· â MCP â Foundry IQ â M365 éæ
â
Level 4: é«çº§çæåŒ AI
â RAG â Fine-tuning â AI Search â Content Understanding
â
Level 3: çæåŒ AI åºç¡
â Model Catalog â Foundry SDK â Prompt Flow â 倿š¡æåºçš
â
Level 2: èªå®ä¹ AI æš¡å
â CLU â QnA â ææ¬åç±» â èªå®ä¹ NER â Custom Vision â Doc Intelligence
â
Level 1: é¢æå»º AI æå¡
â ææ¬åæ â ç¿»è¯ â è¯é³ â åŸååæ â OCR â äººèž â è§é¢ â Doc Intelligence
â
Level 0: åºç¡å¹³å°
Azure AI æå¡å
𿝠â Microsoft Foundry â è®€è¯ææ â åŒåç¯å¢
English Version
1. Overview
AI-102 is the official course for the Microsoft Azure AI Engineer Associate certification. It covers 5 learning paths with 40 modules, spanning: generative AI app development, AI agent building, natural language processing, computer vision, and information extraction.
This article structures ALL AI-102 knowledge from Level 0 (Foundation) to Level 7 (Advanced), building a complete knowledge map where each level builds on the previous one.
Course Overview â 5 Learning Paths
| # | Learning Path | Modules | Core Topics |
|---|---|---|---|
| 1 | Develop Generative AI Apps in Azure | 8 | GenAI, RAG, Fine-tuning, Prompt Flow |
| 2 | Develop AI Agents on Azure | 9 | Agent Service, MCP, Multi-agent, A2A |
| 3 | Develop Natural Language Solutions | 10 | Text Analytics, CLU, Speech, Translation |
| 4 | Develop Computer Vision Solutions | 8 | Image Analysis, OCR, Face, Custom Vision |
| 5 | Develop AI Information Extraction Solutions | 5 | Document Intelligence, AI Search, Content Understanding |
2. Level 0 â Foundation & Setup
ð¯ Goal: Understand the Azure AI ecosystem and set up your development environment
2.1 Azure AI Service Landscape
âââââââââââââââââââââââââââââââââââââââââââââââââââââââ
â Microsoft Foundry â
â (Unified AI Development Platform, formerly AI Studio)â
â â
â ââââââââââââ ââââââââââââ ââââââââââââ â
â â Model â â Prompt â â Agent â â
â â Catalog â â Flow â â Service â â
â ââââââââââââ ââââââââââââ ââââââââââââ â
â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â Azure AI Services â â
â â ââââââââââ ââââââââââ ââââââââââ âââââââââââ â
â â âLanguage â âVision â âSpeech â âDocumentââ â
â â âService â âService â âService â âIntelli.ââ â
â â ââââââââââ ââââââââââ ââââââââââ âââââââââââ â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â Azure OpenAI Service â â
â â GPT-4o â GPT-4 â DALL-E â Whisper â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
â â Azure AI Search â â
â â Index â Skillset â Vector Search â Semantic â â
â ââââââââââââââââââââââââââââââââââââââââââââââââ â
âââââââââââââââââââââââââââââââââââââââââââââââââââââââ
2.2 Key Foundation Concepts
| Concept | Description |
|---|---|
| Microsoft Foundry | Unified AI dev platform (formerly Azure AI Studio) for managing models, data, deployments |
| Azure AI Services | Collection of pre-built cognitive services (Language, Vision, Speech, Decision) |
| Azure OpenAI Service | Access OpenAI models (GPT, DALL-E, Whisper) through Azure |
| Resource & Endpoint | Each AI service requires a resource, accessed via Endpoint + Key or Entra ID |
| REST API & SDK | Two invocation methods: direct HTTP REST calls, or Python/C# SDKs |
2.3 Development Environment Setup
Prerequisites:
âââ Azure subscription (free trial available)
âââ Python 3.8+ or C#/.NET
âââ Visual Studio Code
â âââ Azure AI Foundry extension
â âââ Python / C# extension
âââ Azure CLI
âââ Azure AI Foundry SDK (pip install azure-ai-projects)
2.4 Authentication & Authorization
- API Key: Simple but not recommended for production
- Microsoft Entra ID (formerly AAD): Recommended for production using
DefaultAzureCredential - RBAC Roles:
Cognitive Services User,Cognitive Services Contributor
3. Level 1 â Pre-built AI Services (Out-of-the-Box)
ð¯ Goal: Learn to consume Azureâs pre-built AI capabilities without training models
3.1 Text Analytics
Service: Azure Language Service (Language in Foundry Tools)
| Capability | Description | Use Case |
|---|---|---|
| Sentiment Analysis | Determine positive/negative/neutral sentiment | Customer feedback, social monitoring |
| Key Phrase Extraction | Extract keywords and phrases | Document summarization, tagging |
| Named Entity Recognition | Identify people, places, orgs, dates | Information extraction |
| Language Detection | Identify the language of text | Multi-language routing |
| Entity Linking | Link entities to Wikipedia entries | Knowledge graphs, disambiguation |
| PII Detection | Identify personally identifiable information | Data masking, compliance |
3.2 Translation Service
- Text Translation: Real-time translation across 100+ languages
- Document Translation: Batch translate entire documents with format preservation
- Custom Translator: Train custom translation models with your terminology
- Transliteration: Convert between writing systems
3.3 Speech Services
| Capability | Direction | Description |
|---|---|---|
| Speech-to-Text | Audio â Text | Real-time/batch speech recognition |
| Text-to-Speech | Text â Audio | Natural voice synthesis with custom voices |
| Speech Translation | Audio â Translated Text | Real-time spoken language translation |
| Speaker Recognition | Audio â Identity | Speaker verification and identification |
Key Concepts: Speech Config, Audio Config, SSML (Speech Synthesis Markup Language)
3.4 Image Analysis
Service: Azure Vision Service
| Capability | Description |
|---|---|
| Image Description | Auto-generate natural language descriptions |
| Tag Extraction | Identify objects, scenes, actions |
| Object Detection | Locate objects with bounding boxes |
| Smart Cropping | Auto-crop based on regions of interest |
| Face Detection | Detect face locations and attributes |
| OCR | Extract text from images |
3.5 OCR â Reading Text in Images
- Image Analysis Read API: Simple scenarios, synchronous
- Document Intelligence: Complex documents, asynchronous (see Level 4)
- Supports: printed text, handwriting, multi-language
3.6 Face Detection & Recognition
Three capability tiers:
- Face Detection: Detect face location and attributes (approval required)
- Face Verification: 1:1 comparison â are these the same person?
- Face Identification: 1:N recognition â who is this person?
â ïž Face recognition features are subject to Limited Access policy
3.7 Video Analysis (Video Indexer)
Extracts multi-dimensional insights: face tracking, OCR, speech-to-text, topics, sentiment, scene segmentation, content moderation.
3.8 Prebuilt Document Intelligence Models
| Model | Extracts |
|---|---|
| Invoice | Invoice number, dates, amounts, vendor info |
| Receipt | Merchant, transaction date, line items, totals |
| ID Document | Name, DOB, document number |
| Business Card | Name, title, company, contact info |
| Tax Documents | W-2, 1098 US tax form fields |
4. Level 2 â Custom AI Models (Train Your Own)
ð¯ Goal: Train custom models when pre-built capabilities arenât sufficient
4.1 Conversational Language Understanding (CLU)
| Concept | Description | Example |
|---|---|---|
| Utterance | What the user says | âBook me a flight to Tokyo tomorrowâ |
| Intent | What the user wants to do | BookFlight |
| Entity | Key parameters in the utterance | destination=Tokyo, date=tomorrow |
Training workflow: Define Intents/Entities â Label training data â Train â Evaluate (Precision/Recall/F1) â Deploy â Integrate
4.2 Question Answering
Knowledge-base-powered Q&A system:
- Data Sources: FAQ pages, Word/PDF docs, manual QA pairs
- Multi-turn conversations: Follow-up prompts for guided dialogue
- Precise answers: Extract exact answer spans from passages
- Synonyms: Configure alterations to improve matching
4.3 Custom Text Classification
| Type | Description |
|---|---|
| Single-label | Each document belongs to exactly one category |
| Multi-label | Each document can belong to multiple categories |
Requirements: At least 10 documents/category; recommended 200+
4.4 Custom Named Entity Recognition (NER)
Extract custom entity types from unstructured text:
- Define your entity types (e.g., âProductNameâ, âContractNumberâ)
- Label training data â Train â Evaluate â Deploy
- Metrics: Precision, Recall, F1-Score
4.5 Custom Vision â Image Classification
Training images â Upload â Choose classification type â Train â Evaluate â Publish
âââ Multi-class (one label per image)
âââ Multi-label (multiple labels per image)
Best practices: At least 15 images/class, diverse angles/lighting/backgrounds
4.6 Custom Vision â Object Detection
Locates objects with bounding boxes â not just what but where. Use cases: industrial inspection, retail shelf monitoring, security
4.7 Custom Document Intelligence Models
- Custom template model: Fixed-format forms
- Custom neural model: Semi-structured and unstructured documents
- Composed model: Combine multiple models for different form types
5. Level 3 â Generative AI Fundamentals
ð¯ Goal: Master LLM model selection, deployment, and basic app development
5.1 Model Catalog & Deployment
| Model Provider | Examples | Deployment |
|---|---|---|
| Azure OpenAI | GPT-4o, GPT-4, GPT-3.5-turbo | Standard/Global |
| Meta | Llama 3 | Serverless API / Managed Compute |
| Mistral | Mistral Large, Mixtral | Serverless API |
| Microsoft | Phi-3, Phi-4 | Serverless API / Managed Compute |
| Cohere | Command R+ | Serverless API |
Deployment Types:
- Serverless API (MaaS): Pay-per-token, no infrastructure management
- Managed Compute (MaaP): Dedicated compute, high throughput
- Standard Deployment: Default for Azure OpenAI models
5.2 Microsoft Foundry SDK
Core components:
AIProjectClient: Project-level client for managing resourcesChatCompletionsClient: Chat completion callsEmbeddingsClient: Text embedding generation
5.3 Prompt Flow
Visual LLM application orchestration tool:
- Flow: DAG-based workflow
- Nodes: LLM calls, Python code, tool invocations
- Connections: External service configurations
- Variants: Different prompt versions for A/B testing
- Three Flow Types: Standard, Chat, Evaluation
5.4 Multimodal Generative AI
- Vision-enabled apps: GPT-4o processing images (descriptions, visual Q&A, chart interpretation)
- Speech-enabled apps: STT â LLM â TTS pipeline for voice assistants
- Image generation: DALL-E for text-to-image, editing, and variations
6. Level 4 â Advanced Generative AI
ð¯ Goal: Master RAG, Fine-tuning, and enterprise knowledge retrieval
6.1 RAG â Retrieval Augmented Generation
Why RAG? LLMs have training cutoff dates, donât know your private data, and prompts have token limits.
RAG Architecture:
Your Data â Chunking â Embedding â Azure AI Search Index
â
User Question â Retrieve relevant chunks â Combine with prompt â LLM â Grounded answer
Key Steps:
- Data preparation: Document chunking & embedding generation
- Index creation: Azure AI Search vector index
- Retrieval: Hybrid search (keyword + vector + semantic ranking)
- Prompt engineering: System prompt guiding model to use retrieved context
- Response generation: LLM generates grounded answers
6.2 Fine-tuning
| Scenario | RAG | Fine-tuning |
|---|---|---|
| Need latest knowledge | â Preferred | â Not suitable |
| Change model behavior/style | â Limited | â Preferred |
| Learn specific formats | â | â |
| Reduce token consumption | â | â |
| Quick implementation | â | â Takes time |
Process: Prepare JSONL training data â Upload â Select base model â Configure hyperparameters â Train â Evaluate â Deploy
6.3 Azure AI Search â Knowledge Mining
Three capabilities: Data Ingestion â AI Enrichment (Skillsets) â Intelligent Search
Built-in Skills: OCR, image analysis, key phrase extraction, entity recognition, language detection, translation, document cracking, text merge, custom skills (call your own API)
6.4 Azure Content Understanding
Next-generation multimodal content analysis:
- Documents, images, audio, video â unified processing
- Relationship to Document Intelligence: Content Understanding is newer and more unified
7. Level 5 â AI Agent Development
ð¯ Goal: Build autonomous agents that can plan and execute tasks
7.1 Core Agent Concepts
Agent = LLM + Tools + Memory + Planning
| Component | Purpose |
|---|---|
| LLM (Brain) | Understanding, reasoning, decision-making |
| Tools | Search, code execution, API calls, file operations |
| Memory | Short-term (conversation) and long-term (persistent) |
| Planning | Decompose complex tasks into steps |
7.2 Microsoft Foundry Agent Service
Built-in tools:
- Code Interpreter: Run Python, analyze data, generate charts
- File Search: Search uploaded file contents
- Bing Grounding: Real-time web information
- Azure AI Search: Enterprise knowledge base search
7.3 Custom Tool Integration
Define function schemas for your own tools; the agent calls them when needed and you implement the actual logic.
7.4 MCP Tools (Model Context Protocol)
Open standard protocol for dynamic tool discovery and invocation:
Agent ââ MCP Client ââ MCP Server ââ External Tools/Services
7.5 Foundry IQ â Knowledge-Enhanced Agents
Shared knowledge platform that multiple agents can access:
- Data optimization for retrieval quality
- Agent instruction configuration for consistent, cited responses
7.6 Microsoft 365 Integration
Publish agents to Teams, M365 Copilot, and access workplace data through Work IQ.
8. Level 6 â Multi-Agent & Advanced Orchestration
ð¯ Goal: Build multi-agent collaborative systems and enterprise AI workflows
8.1 Microsoft Agent Framework (Semantic Kernel)
Core concepts:
- Kernel: AI orchestration engine
- Plugins: Reusable capability modules
- Planner: Auto-decompose goals into steps
- Memory: Vector storage and semantic memory
8.2 Multi-Agent Orchestration
Orchestration Strategies:
| Strategy | Description | Use Case |
|---|---|---|
| Sequential | Fixed order rotation | Simple pipelines |
| Round-Robin | Cyclic rotation | Equal participation |
| Selection Strategy | LLM decides next speaker | Flexible collaboration |
| Termination Strategy | Define completion conditions | Auto-stop |
8.3 A2A Protocol (Agent-to-Agent)
Cross-platform agent communication protocol:
- Agent Discovery: Discover remote agent capabilities
- Direct Communication: Agent-to-agent messaging
- Task Coordination: Cross-agent task orchestration
8.4 Agent-Driven Workflows
Build intelligent workflows with Microsoft Foundry:
- Orchestrate multiple agents and components
- Conditional branching and loops
- Error handling and retries
- Human-in-the-loop review nodes
9. Level 7 â Responsible AI & Production
ð¯ Goal: Ensure AI solutions are safe, reliable, and responsible
9.1 Responsible AI Principles
| Principle | Description |
|---|---|
| Fairness | AI should treat all people fairly |
| Reliability & Safety | AI should perform reliably without causing harm |
| Privacy & Security | Protect user data and privacy |
| Inclusiveness | AI should empower everyone |
| Transparency | AI decisions should be understandable |
| Accountability | Humans should be accountable for AI systems |
9.2 Content Safety for Generative AI
Azure AI Content Safety provides input/output filtering:
User Input â Input Filter â LLM â Output Filter â Response
â â
Detect: Detect:
- Hate - Hate
- Violence - Violence
- Sexual - Sexual
- Self-harm - Self-harm
- Jailbreak - Misinformation
Four severity levels: Safe â Low â Medium â High
9.3 Evaluating Generative AI Applications
| Metric | Evaluates |
|---|---|
| Groundedness | Is the answer based on provided context? |
| Relevance | Is the answer relevant to the question? |
| Coherence | Is the answer logically coherent? |
| Fluency | Is the language natural and fluent? |
| Similarity | How similar to the reference answer? |
| F1 Score | Token overlap with reference answer |
Evaluation methods: Manual scoring, AI-assisted (LLM-as-judge), Automated metrics via Prompt Flow evaluation flows
10. Complete Knowledge Map
Level 7: Responsible AI & Production
â Responsible AI Principles â Content Safety â Evaluation Metrics
â
Level 6: Multi-Agent & Advanced Orchestration
â Semantic Kernel â Multi-Agent Chat â A2A Protocol â Agent Workflows
â
Level 5: AI Agent Development
â Foundry Agent Service â Custom Tools â MCP â Foundry IQ â M365 Integration
â
Level 4: Advanced Generative AI
â RAG â Fine-tuning â AI Search â Content Understanding
â
Level 3: Generative AI Fundamentals
â Model Catalog â Foundry SDK â Prompt Flow â Multimodal Apps
â
Level 2: Custom AI Models
â CLU â Q&A â Text Classification â Custom NER â Custom Vision â Doc Intelligence
â
Level 1: Pre-built AI Services
â Text Analytics â Translation â Speech â Image Analysis â OCR â Face â Video â Doc Intelligence
â
Level 0: Foundation
Azure AI Landscape â Microsoft Foundry â Auth & Security â Dev Environment
8. åèèµæ (References)
- AI-102 Course Page â Official course landing page
- Develop Generative AI Apps Learning Path â Path 1: GenAI development
- Develop AI Agents on Azure Learning Path â Path 2: AI Agent development
- Develop Natural Language Solutions Learning Path â Path 3: NLP solutions
- Develop Computer Vision Solutions Learning Path â Path 4: Computer Vision
- Develop AI Information Extraction Solutions Learning Path â Path 5: Information extraction
- Azure AI Engineer Associate Certification â Certification details