Deploying Amazon AI Services for Inference

AWS Certified Machine Learning Engineer – Associate · Deployment and Orchestration of ML Workflows · By Super Admin

AWS AI services expose pretrained models only through boto3 API calls, each with distinct client methods for vision, document, speech, language, conversational, and recommendation tasks.

Although Amazon SageMaker AI is often used to build, train, and deploy custom ML models, it also integrates with several fully managed AI services provided by AWS that offer pretrained models incorporated into applications via API calls.

Consuming Pretrained Models via API

These pretrained models are hosted in managed AWS accounts owned by the respective services, so there is limited access to them; the models are exposed for consumption only through API calls, typically made using the AWS SDK for Python (boto3). Amazon SageMaker AI can be used to orchestrate these API calls within ML workflows, allowing scalable processing without needing direct access to the underlying models.

Vision, Document, and Speech Services

Amazon Rekognition offers pretrained models for image and video analysis; a client is created with boto3.client('rekognition'), and the detect_labels method analyzes images stored in Amazon S3 to identify objects, scenes, and activities, returning labels with confidence scores. Amazon Textract provides pretrained models for extracting text and data from scanned documents, with key methods including analyze_document, detect_document_text, and get_document_analysis. Amazon Polly converts text into lifelike speech using pretrained models, with key methods including synthesize_speech, describe_voices, and list_lexicons. Amazon Transcribe offers pretrained models for speech-to-text conversion, with key methods including start_transcription_job, get_transcription_job, and list_transcription_jobs for managing transcription tasks.

Language, Conversational, and Recommendation Services

Amazon Comprehend offers pretrained models for natural language processing tasks such as sentiment analysis, entity recognition, and key phrase extraction, with key methods including detect_sentiment, detect_entities, and detect_key_phrases. Amazon Lex provides pretrained models for building conversational interfaces using voice and text; its client is created with boto3.client('lexv2-runtime'), and the key method recognize_text manages interactions with a chatbot. Amazon Personalize offers pretrained models for creating individualized recommendations, with key methods including get_recommendations, get_personalized_ranking, and create_campaign for managing recommendation tasks.

Key takeaways

  • AI services host pretrained models in AWS-managed accounts that are accessible only through API calls.
  • Amazon SageMaker AI can orchestrate calls to AI services within broader ML workflows.
  • Each service exposes specific key methods, such as detect_labels for Rekognition, synthesize_speech for Polly, and get_recommendations for Personalize.
0.0 (0 ratings)

Discussion (0)

No discussion on this article yet — be the first to comment.

Please sign in to join the discussion.

Deploying Amazon AI Services for Inference | SecViva