SageMaker
💡 Definition
Amazon SageMaker is a fully managed service that provides every developer and data scientist with the ability to build, train, and deploy machine learning (ML) models quickly.
🔑 Key Concepts
- End-to-End ML: Covers the entire ML workflow: labeling data, building models, training, tuning, and deploying.
- SageMaker Studio: A web-based IDE for all ML steps.
- Managed Infrastructure: AWS manages the underlying instances for training and hosting models.
- Algorithms: Includes built-in high-performance algorithms (e.g., XGBoost).
⚙️ How it Works
- Build: Use Jupyter notebooks to write code.
- Train: SageMaker spins up a training cluster, trains the model on your data (in S3), and saves the model artifacts.
- Deploy: SageMaker deploys the model to a hosting endpoint for real-time predictions.
🎯 Use Cases
- Custom Models: When pre-trained AI services (like Rekognition) aren't specific enough.
- Data Science: Analyzing large datasets and creating predictive models.
- ML Ops: Automating the ML lifecycle.
💰 Pricing Model
- Instance Hours: Charged for the compute instances used for building (notebooks), training, and hosting.
- Storage: Charged for storage used.
📝 Exam Tips (CLF-C02)
- For "building, training, and deploying" your own custom models.
- Targeted at developers and data scientists.
- It's a platform/framework, not a simple API call like the AI services.
See Also: * Rekognition * Comprehend