AI/ML on AWS: An Overview
Both AI/ML and traditional data analytics need clean and accessible data in a format that's usable by analytics tools and AI algorithms. AWS offers a comprehensive AI/ML stack to address a wide range of business use cases, from predicting trends and making decisions to detecting anomalies.
Common ML Business Use Cases​
ML models can power diverse applications and solve complex business problems, including:
- Predict Trends: Forecast future stock prices, sales demand, or customer churn.
- Make Decisions: Route callers to the right department, optimize logistics, or personalize content.
- Detect Anomalies: Identify bank fraud, unusual network activity, or equipment failures.
AWS AI/ML Solutions: The Three Tiers​
The AWS AI/ML stack is composed of three tiers, offering varying levels of control and customization:
| Tier | Description | Primary Services/Approach |
|---|---|---|
| Tier 1: AI Services | - Pre-built models already trained to perform specific functions. - These are ready-to-use, managed services for quick solutions without ML expertise. | - Amazon Comprehend. - Amazon Polly. - Amazon Transcribe. - Amazon Translate. - Amazon Kendra. - Amazon Rekognition. - Amazon Textract. - Amazon Lex. - Amazon Personalize. |
| Tier 2: ML Services | - A more customized approach where you build, train, and deploy your own ML models with fully managed infrastructure. - Offers more control over ML solutions without managing underlying infrastructure. | - Amazon SageMaker. |
| Tier 3: ML Frameworks and Infrastructure | - A completely custom approach for organizations with highly specialized needs. - Provides complete control over the ML training process using purpose-built chips and popular ML frameworks. | - PyTorch. - TensorFlow. - Apache MXNet on EC2, EMR, ECS, etc. - (underlying infrastructure). |
This tiered approach allows you to choose the right level of abstraction for your specific needs, from quick integration of pre-trained models to deep customization of ML frameworks.
Generative AI on AWS​
Within the AI/ML stack, AWS offers specialized solutions for Generative AI, enabling the creation of new content and accelerating productivity:
| Service Name | Key Focus |
|---|---|
| Amazon SageMaker JumpStart | ML hub with FMs and pre-built ML solutions for rapid deployment and fine-tuning. |
| Amazon Bedrock | Fully managed service for adapting and deploying FMs from Amazon and leading AI companies via a single API. |
| Amazon Q | Interactive AI assistant integrated with company information repositories for business and developer productivity. |