Link copied to clipboard!
Group 13: AI & Machine Learning
Intelligent services and ML platform capabilities: SageMaker, Comprehend, Rekognition, Polly, Lex.
AI/ML Principle: Match AI service complexity to expertise level, data characteristics, and application requirements - from pre-trained APIs to custom model development.
Services & Roles
SageMaker
End-to-end ML platform for building, training, and deploying custom models.
- Jupyter notebooks
- Model training
- Endpoint deployment
Comprehend
Natural language processing for sentiment analysis and entity extraction.
- Sentiment analysis
- Entity extraction
- Topic modeling
Rekognition
Computer vision for image and video analysis with facial recognition.
- Object detection
- Facial analysis
- Content moderation
Polly
Text-to-speech service with neural voices and SSML support.
- Neural voices
- SSML support
- Speech marks
Lex
Conversational AI for building chatbots and voice interfaces.
- Intent recognition
- Slot filling
- Multi-turn conversations
Key Differences
| Dimension | SageMaker | Comprehend | Rekognition | Polly | Lex |
|---|---|---|---|---|---|
| Complexity | Full ML platform | Pre-trained NLP | Pre-trained vision | Pre-trained TTS | Pre-trained conversational |
| Customization | Full control | Custom models | Custom labels | Voice selection | Custom intents |
| Expertise Required | High | Medium | Low | Low | Medium |
| Use Case | Custom ML models | Text analysis | Image analysis | Voice synthesis | Chatbots |
| Data Input | Any structured data | Text documents | Images/videos | Text | Speech/text |
Selection Model
0–10 sliders weight AI/ML requirements, expertise level, and application-specific needs.
Score_SageMaker = 0.32*C_customModels + 0.26*C_mlExpertise + 0.18*C_endToEndPlatform + 0.12*C_modelTraining + 0.08*C_experimentTracking + 0.04*C_jupyterNotebooks
Score_Comprehend = 0.30*C_textAnalysis + 0.26*C_sentimentAnalysis + 0.18*C_entityExtraction + 0.12*C_nlpRequirements + 0.08*C_documentProcessing + 0.06*C_topicModeling
Score_Rekognition= 0.32*C_imageAnalysis + 0.28*C_computerVision + 0.18*C_facialRecognition + 0.12*C_objectDetection + 0.06*C_contentModeration + 0.04*C_videoAnalysis
Score_Polly = 0.34*C_textToSpeech + 0.30*C_voiceSynthesis + 0.18*C_neuralVoices + 0.12*C_speechMarks + 0.04*C_ssmlSupport + 0.02*(10 - C_customModels)
Score_Lex = 0.32*C_chatbots + 0.28*C_conversationalAI + 0.18*C_intentRecognition + 0.12*C_voiceInterfaces + 0.06*C_slotFilling + 0.04*C_multiTurnConversations
SageMaker {{vm.scores.sagemaker|number:2}}
Comprehend {{vm.scores.comprehend|number:2}}
Rekognition {{vm.scores.rekognition|number:2}}
Polly {{vm.scores.polly|number:2}}
Lex {{vm.scores.lex|number:2}}
Primary Emphasis: {{vm.recommended.name}} ({{vm.recommended.score|number:2}})
Heuristics
- High C_customModels + C_mlExpertise → SageMaker for bespoke ML solutions.
- Strong C_textAnalysis + C_sentimentAnalysis → Comprehend for NLP tasks.
- High C_imageAnalysis + C_computerVision → Rekognition for vision applications.
- C_textToSpeech + C_voiceSynthesis → Polly for speech generation.
- C_chatbots + C_conversationalAI → Lex for conversational interfaces.
Anti-Patterns
- Using SageMaker for simple pre-trained model use cases (overkill).
- Using Comprehend for image analysis (use Rekognition instead).
- Using Rekognition for text analysis (use Comprehend instead).
- Building custom TTS when Polly meets requirements.
Summary
Choose SageMaker for custom ML, Comprehend for text, Rekognition for vision, Polly for speech, Lex for conversations.