AIF-C01: SageMaker Ecosystem

Amazon SageMaker

Fully managed ML platform; end-to-end service to build; train; tune; deploy; and manage machine learning models at scale with managed infrastructure

SageMaker Canvas

No-Code Visual; drag-and-drop ML; designed for Business Analysts

SageMaker JumpStart

Pre-trained Model Hub; one-click deployment of Foundation Models & open-source LLMs; hosted on dedicated SageMaker endpoints with full infrastructure & hyperparameter control

SageMaker Autopilot

AutoML; automated model building & algorithm selection; tabular CSV data

SageMaker Ground Truth

Data Labeling; human-in-the-loop (HITL) annotation; raw dataset marking

SageMaker Clarify

Bias Detection & Model Explainability; checks fairness metrics & feature importance

SageMaker Feature Store

Centralized ML Feature Repository; stores and reuses curated tabular features; guarantees online/offline consistency

SageMaker Data Wrangler

Visual Data Preparation; low-code cleaning; transformation; & feature engineering; exports to SageMaker Pipelines

SageMaker Model Cards

AI Governance & Documentation; single source of truth for model metadata; intended use; metrics; and ethical considerations

SageMaker Model Registry

Model Versioning & Cataloging; tracks model lineage; manages approval workflows (Pending -> Approved); and triggers pipelines

SageMaker Endpoints

Managed hosting for model inference; exposes trained ML models via REST APIs; supports Real-time (low latency); Serverless (intermittent traffic); Async (large payloads/long processing); and Batch Transform (offline bulk data)

SageMaker Experiments

Iterative ML trial tracking; logs; organizes; and compares hyperparameters; metrics; and data artifacts across training runs

SageMaker Pipelines

Automated ML Workflows; CI/CD pipeline orchestration; manages DAGs for repeatable model training; evaluation; and deployment

SageMaker Model Monitor

Production Endpoint Monitoring; detects real-time data drift; concept drift; bias drift; and model quality degradation over time

SageMaker Role Manager

Access Governance; provides baseline IAM role templates and pre-built permission sets for specific ML personas (data scientists / MLOps)