Before building a custom AI model, check if there’s an existing service available. AWS offers pre-trained AI services accessible via APIs, saving you time and resources.
Amazon Rekognition (Computer Vision)
- Pre-trained deep learning service for image and video analysis (including streaming videos).
- Face Recognition: Can verify identity by comparing faces to reference images (e.g., ID cards, employee badges).
- Object Detection: Recognizes objects in images/videos and can label them for searchability.
- Custom Object Recognition: Can be trained on specific images to identify proprietary objects.
- Content Moderation: Detects inappropriate, explicit, or violent content in images and videos.
Example:
- Upload a reference image of a face, and Amazon Rekognition can match it to other images/videos with high confidence (e.g., 99.8%).
Amazon Textract (Text Extraction)
- Extracts text: Reads scanned documents, forms, tables, and handwriting.
- Beyond OCR: It can handle structured data from documents like invoices or forms.
Amazon Comprehend (Natural Language Processing)
- Sentiment Analysis: Analyzes customer feedback or comments to determine sentiment (positive, negative, neutral).
- Text Insights: Discover relationships and insights in text data.
- Detect PII (Personally Identifiable Information): Finds sensitive information in text (e.g., names, addresses, emails, phone numbers).
Example:
- Amazon Comprehend can scan an email to find PII, like credit card numbers or names, and return a confidence score for each entity.
Amazon Lex (Voice and Text Interfaces)
- Build chatbots and interactive voice response systems.
- Uses the same technology as Amazon Alexa.
- Common use cases: Customer service chatbots and call center systems.
Amazon Transcribe (Speech Recognition)
- Automatic Speech Recognition (ASR): Converts speech to text.
- Supports over 100 languages.
- Common use case: Captioning live audio or video streams in real time.
Summary of More AWS AI Services
Amazon Polly
- Converts text into natural-sounding speech in many languages using deep learning.
- Use cases include converting articles to speech and guiding callers in interactive voice systems.
- Helps companies engage users, especially for visually impaired customers.
Amazon Kendra
- Performs intelligent search using machine learning.
- Understands natural language questions and returns relevant results, such as “How do I connect my Echo Plus to my network?”
Amazon Personalize
- Generates personalized recommendations for customers in sectors like retail and media.
- Example: Suggesting products like “you might also like” in an e-commerce app or segmenting customers for better marketing.
Amazon Translate
- Provides fluently accurate translations between 75 languages.
- Built on neural networks for context-aware translations, ideal for real-time translation in online chat.
Amazon Forecast
- AI service for time series forecasting.
- Use cases include predicting future trends in sales, inventory, or healthcare by analyzing historical data.
Amazon Fraud Detector
- Identifies fraudulent online activities (payment fraud, fake accounts, etc.).
- Uses pre-trained models to detect fraud in transactions, product reviews, and account management.
Amazon Bedrock
- Fully managed service for building generative AI applications.
- Offers foundation models that can be customized using your own training data.
- Supports Retrieval Augmented Generation (RAG) for knowledge retrieval outside the model’s training data.
Amazon SageMaker
Features pre-trained models to accelerate the process of building and training models.
Offers custom ML models and workflows beyond pre-built AI services.
Includes tools for data preparation, large-scale model training, deployment, and real-time inference.
Key Takeaways
- AWS AI Services: Pre-trained models for various use cases like computer vision, natural language processing, speech recognition, and more.
- Cost-effective Solution: Consider using existing AWS services rather than building your own custom model.
