Machine Learning

Amazon Rekognition is a managed image and video analysis service using deep learning — face detection, object detection, content moderation, celebrity recognition, text-in-image detection.

Amazon Polly is a text-to-speech service. Provides natural-sounding voices in many languages — used for IVR, narrating content, accessibility.

Amazon Transcribe is the AWS automatic speech recognition (ASR) service. Converts audio to text — supports streaming, custom vocabularies, speaker diarization, automatic language detection.

Amazon Translate provides neural machine translation between many languages. Real-time and batch translation; custom terminology.

Amazon Comprehend is a natural language processing (NLP) service. Detects entities, key phrases, sentiment, topics, syntax, and PII in text. Comprehend Medical for medical text.

Amazon SageMaker is a fully managed ML platform — covers the entire ML lifecycle: data labeling, notebooks, training, hyperparameter tuning, model deployment, monitoring.

Amazon Lex is the service powering Alexa — builds chatbots with automatic speech recognition (ASR) and natural language understanding (NLU). Integrates with Connect for IVR.

Amazon Connect is a fully managed cloud contact center — replaces traditional call centers. Integrates with Lex for chatbots, Polly for TTS, and many other AWS services.

Amazon Textract uses ML to automatically extract printed text, handwriting, forms, and tables from scanned documents — far beyond basic OCR.

Amazon Personalize provides real-time personalized recommendations (similar to those used at Amazon.com) — same ML technology as Amazon's own recommendation engine.

Amazon Forecast is a managed time-series forecasting service — uses the same technology Amazon uses for retail demand forecasting. (Tip: AWS now also encourages using Amazon SageMaker Canvas for new forecasting workloads since it includes time-series forecasting in a no-code experience.)

Amazon Bedrock provides serverless access to multiple foundation models from leading AI companies (Anthropic, AI21, Cohere, Meta, Stability AI, Amazon) through a single API for generative AI apps.

Amazon Q is AWS's generative AI assistant — Amazon Q Developer (formerly CodeWhisperer) for code, Amazon Q Business (knowledge worker assistant connecting to enterprise data), and Amazon Q in QuickSight for BI.

Amazon Fraud Detector is a managed service that uses ML to detect online fraud — fake accounts, payment fraud, identity fraud — based on Amazon's 20+ years of fraud-detection experience.

Amazon DevOps Guru uses ML to detect operational anomalies (latency spikes, error spikes, resource exhaustion) and recommend remediations. It analyzes CloudWatch metrics, AWS X-Ray traces, AWS Config events, and CloudTrail logs.

AWS has multiple healthcare/medical AI services: Comprehend Medical (NLP for medical text), Transcribe Medical (medical speech-to-text), HealthLake (clinical data store), and HealthOmics (genomics).

Amazon SageMaker Canvas provides a visual, no-code interface for building ML models — designed for business analysts without coding experience. Connects to many data sources.

SageMaker Autopilot is automated ML (AutoML) — automatically explores and tunes models, lets you see the steps it took (transparent AutoML).

Amazon SageMaker Studio is a web-based ML IDE — provides notebooks, debugging, model training, deployment, and monitoring in a single interface.

Contact Lens for Amazon Connect uses ML to analyze call center conversations — sentiment, talk speed, interruptions, themes — helping improve customer experience and agent performance.

Amazon Bedrock is serverless, offers multiple foundation models (Anthropic, AI21, Cohere, Meta, Stability AI, Amazon Titan), supports knowledge bases, agents, and guardrails for responsible AI.

P-family (P3 with V100, P4 with A100, P5 with H100) and Trainium (Trn1) instances are optimized for ML training. For inference: G-family and Inferentia (Inf1, Inf2).

Amazon Rekognition Custom Labels lets customers train custom computer-vision models on their own images (e.g., detect their company's products, brand logos, machine parts) without ML expertise — backed by Rekognition's transfer-learning pipeline.

Amazon Transcribe Medical is a speech-to-text service tuned for medical terminology — used for clinical documentation, telehealth, and patient-provider conversations. (Comprehend Medical is NLP for medical text; Transcribe Medical is for audio.)

Amazon OpenSearch Service (formerly Elasticsearch Service) is a managed search and analytics service — used for log analytics, full-text search, observability dashboards.

Amazon Kendra is an ML-powered enterprise intelligent search service designed to answer natural-language questions and return precise answers from unstructured documents (PDFs, HTML, SharePoint, ServiceNow, etc.). OpenSearch is a search/log-analytics engine but does not provide natural-language Q&A out of the box. Comprehend extracts entities/sentiment, and Translate is for language translation.

AWS forked Elasticsearch and Kibana in 2021 and rebranded the managed service from Amazon Elasticsearch Service to Amazon OpenSearch Service (with OpenSearch Dashboards replacing Kibana). Older study materials may still reference the old name — they are the same service.

Amazon Q is AWS's generative-AI assistant family. Q Business answers questions over your enterprise data sources; Q Developer (formerly CodeWhisperer) helps write and review code in IDEs; Q in QuickSight generates dashboards and natural-language answers from your data. Lex builds chatbots, SageMaker is for building/training ML models, and Bedrock provides foundation-model APIs but is not the assistant itself.