# Choose a service

Pick the service that matches how you want to work. Every guide is self-contained and reproducible.

  
    SageMaker SDK
    Full control from Python: deploy any Hub model to a managed endpoint, or run custom training jobs on managed infrastructure.
    Quickstart →
  
  
    JumpStart
    A curated model catalog with performant defaults, deployable in a few clicks from SageMaker Studio.
    Quickstart →
  
  
    Bedrock
    JumpStart models behind the managed Bedrock APIs: Agents, Knowledge Bases, Guardrails, and Model Evaluations.
    Quickstart →
  
  
    EC2, ECS, and EKS
    Run the Deep Learning Containers directly on AWS compute services when you need control over networking and orchestration.
    Quickstart →
  
  
    Inference Endpoints
    Hugging Face manages the infrastructure for you, optimized for cost and throughput.
    Guide →
  

All offerings run the same Hugging Face [Deep Learning Containers](./dlcs). For end-to-end recipes (TRL fine-tuning, embedding models, Inferentia2, and more), see the **Examples** under SageMaker SDK in the sidebar.