Amazon SageMaker
Amazon SageMaker is a fully managed machine learning platform that enables developers and data scientists to build, train, and deploy machine learning models at scale. SageMaker removes the heavy lifting from each step of the machine learning process, providing built-in algorithms, managed Jupyter notebooks, distributed training, automatic model tuning, and one-click deployment to production endpoints with auto-scaling.
APIs
Amazon SageMaker API
The Amazon SageMaker control plane API for creating and managing SageMaker resources including notebook instances, training jobs, models, endpoints, pipelines, experiments, feat...
Amazon SageMaker Runtime API
The Amazon SageMaker AI runtime API for invoking deployed model endpoints to get real-time inference predictions.
Amazon SageMaker Feature Store Runtime API
Data plane API operations for the Amazon SageMaker Feature Store supporting put, delete, and retrieve operations for ML features.
Amazon SageMaker Metrics Service API
Data plane API operations for Amazon SageMaker Metrics for putting and retrieving metrics related to training runs.
Amazon SageMaker Geospatial API
APIs for creating and managing Amazon SageMaker geospatial capabilities including earth observation jobs and vector enrichment jobs.
Amazon SageMaker Edge Manager API
SageMaker Edge Manager dataplane service for communicating with active edge agents running ML models on edge devices.
Capabilities
Amazon SageMaker ML Lifecycle Management
Unified capability for managing the end-to-end machine learning lifecycle including notebook development, training, model management, and endpoint deployment. Used by ML Enginee...
Run with NaftikoFeatures
Fully integrated development environment for ML work with notebooks, debugging, and experiment tracking.
Purpose-built infrastructure for distributed training that reduces foundation model training time by up to 40%.
Hub providing access to foundation models, pre-built algorithms, and one-click deployment.
Automated model creation with complete visibility and transparency.
No-code visual interface for creating ML models without writing code.
Store, share, and manage features for machine learning models.
Data preparation tool that reduces transformation workflow time significantly.
Incorporates human feedback throughout the ML lifecycle for data labeling.
Purpose-built CI/CD service for machine learning workflows.
Automatically detects concept drift and data quality issues in deployed models.
Provides machine learning explainability and bias detection.
Streamlines tracking and management of ML experiments.
Access controls and transparency across the full ML lifecycle with audit trails.
Use Cases
Build custom generative AI applications using proprietary data with foundation model fine-tuning.
Train and deploy ML models across the entire machine learning lifecycle from exploration to production.
Query and analyze data across unified sources with built-in SQL analytics and data processing.
Manage data and AI artifacts with fine-grained security controls and compliance tooling.
Build and deploy computer vision models for image classification, object detection, and segmentation.
Train and deploy NLP models for text classification, entity recognition, and language generation.
Build real-time fraud detection models with low-latency inference endpoints.
Deploy ML models on edge devices for predictive maintenance use cases.
Integrations
Store training data, model artifacts, and inference outputs in Amazon S3 data lakes.
Zero-ETL integration for near real-time data ingestion from Redshift warehouses.
Store and manage Docker containers for custom training and inference environments.
Trigger ML inference pipelines and post-processing workflows with Lambda functions.
Trigger SageMaker pipelines and workflows based on events.
Orchestrate multi-step ML workflows using Step Functions state machines.
Lakehouse architecture supporting Apache Iceberg-compatible data tools.
SageMaker Catalog built on Amazon DataZone for data discovery and governance.
Natural language assistance integrated into SageMaker Unified Studio.
Deploy Hugging Face models directly via SageMaker JumpStart.