Azure AI Engineer AI-102 Course

A structured Azure AI Engineer AI-102 course covering AI planning, computer vision, language, speech, Azure OpenAI, deployment and monitoring, with linked practice questions.

Level: Azure AI Engineer Associate Difficulty: advanced 5 lessons 60 min
Course progress 0 / 5
Back to courses

What you will learn

  • Describe AI solution planning and responsible AI.
  • Explain computer vision services.
  • Describe language and document intelligence.
  • Explain speech and conversational AI.
  • Describe Azure OpenAI, deployment and monitoring.

Before you start

  • Basic Azure knowledge is helpful.
  • Some experience with AI or data concepts is recommended.
  • No previous Azure certification is required.

Lesson 1 AI Solution Planning and Responsible AI

Azure AI Engineer AI-102 validates building AI solutions with Azure AI services. Azure AI Foundry is a unified platform for projects, hubs, models, and endpoints. Azure AI Services provide prebuilt APIs for vision, language, speech, and decision. Responsible AI principles include fairness, reliability, privacy, transparency, and accountability. Content Safety filters harmful content. Engineers plan resources, keys, endpoints, quotas, and responsible AI practices before building. Engineers also choose the right pricing tier, monitor quota usage, and document decisions for compliance.

Example

Example: A team creates an AI hub, adds a project, enables Content Safety, and reviews responsible AI guidance before deploying a language model.

Lesson 2 Computer Vision

Computer vision solutions analyze images and video. Azure AI Vision provides image analysis, OCR, captions, thumbnails, and face detection. Custom Vision builds custom image classification and object detection models. OCR extracts text from images, object detection draws bounding boxes, and face detection locates faces. Engineers choose the right service and train custom models when prebuilt APIs are not enough. Image analysis can also return adult content flags and smart-cropped thumbnails for product catalogs.

Example

Example: A document app uses OCR to extract invoice text and Azure AI Vision to generate a caption for a product photo.

Lesson 3 Language and Document Intelligence

Language solutions understand and generate text. Azure AI Language provides sentiment analysis, key phrase extraction, entity recognition, PII detection, and custom question answering. Translator converts text between languages, and conversational language understanding extracts intents and entities. Document Intelligence extracts structured data from forms and documents. Engineers use these services to process customer feedback, support tickets, and business documents. Custom language models can be trained to recognize domain-specific terms and improve accuracy.

Example

Example: A support system uses Language to detect sentiment, PII detection to redact personal data, and Document Intelligence to extract order numbers from emails.

Lesson 4 Speech and Conversational AI

Speech solutions convert audio to text, text to audio, and translate spoken language. Azure AI Speech provides speech-to-text, text-to-speech, custom speech, speaker recognition, pronunciation assessment, and real-time transcription. SSML controls voice and prosody. Engineers build conversational AI with speech endpoints, custom models, and language understanding to serve multilingual users. Batch transcription can process recorded calls, while streaming APIs handle live audio with lower latency.

Example

Example: A customer service app uses speech-to-text for live calls, custom speech for product terms, and text-to-speech to reply in the caller language.

Lesson 5 Azure OpenAI and AI Operations

Azure OpenAI hosts GPT models, embeddings, and content filtering. Prompts guide model behavior, completions are generated text, tokens are text units, and temperature controls randomness. Engineers deploy models to managed endpoints, use API keys for authentication, monitor with Azure Monitor and Application Insights, and evaluate output quality. Containers provide portable deployment, quotas limit usage, and versioning helps manage model changes. Teams should also define fallback behavior, track API versions, and set up alerts for error rates and latency targets.

Example

Example: A team deploys a GPT model, uses a system message to set assistant behavior, monitors latency with Azure Monitor, and evaluates responses before release.