Stop 16 · AI APIs & Cloud AI PlatformsTopics · 6/9

AWS Bedrock (Overview)

In short

If your company already runs on AWS, managers will ask: can we use LLMs without sending data to yet another vendor? Amazon Bedrock is AWS's answer.

AWS's managed service for accessing multiple foundation models (Claude, Llama and others) through a single API, with AWS-native security and billing.

The blob holds a tray and chooses between several colourful food stalls in a mall food court.

Think of it like this

The mall food court

A food court has many stalls, dosa, biryani, pizza, but one security gate and one card you pay with. Bedrock is that: models from several companies inside your AWS account, with AWS logins, AWS billing and AWS rules on who can use what.

Watch it happen

A Converse call through Bedrock

Your app

AWS IAM

Bedrock

converse(modelId, ...)

1/5Your app calls the bedrock-runtime Converse API, signed with AWS credentials, not a vendor API key.

The big ideas

What you need to know

Many models, one API

Bedrock hosts models from Amazon and partner companies. The Converse API gives them one common request shape.

AWS security built in

Access is controlled with IAM roles, calls are logged in CloudTrail, and traffic can stay inside your VPC.

Regions matter

Not every model is in every region. Check model availability for your region before you build.

More than chat

Bedrock also offers Knowledge Bases for RAG, Agents and Guardrails as managed features.

See it in code

Walk through the code

python

1import boto32 3bedrock = boto3.client("bedrock-runtime", region_name="us-east-1")4 5response = bedrock.converse(6    modelId="amazon.nova-lite-v1:0",  # pick an ID from your Bedrock console7    messages=[{"role": "user", "content": [{"text": "Explain RAG in one line."}]}],8    inferenceConfig={"maxTokens": 300},9)10print(response["output"]["message"]["content"][0]["text"])11print(response["usage"])

1/3A bedrock-runtime client in one region. Auth comes from your AWS profile or role.

The Converse API with boto3. Uses your normal AWS credentials. Swap modelId to try another model, the rest stays the same.

Save yourself the pain

Common mistakes

Oops

Getting AccessDenied and assuming the code is wrong.

Do this instead

Check two things: model access is enabled in the Bedrock console, and your IAM role allows bedrock:InvokeModel.

Oops

Leaving test resources running on a personal AWS account.

Do this instead

Set an AWS budget alert before experimenting, and delete anything you created when done.

Your mission

Explore Bedrock safely

~30 min

Quick check

Why might a bank choose Bedrock over calling a model vendor's API directly?

Go deeper

1Read Bedrock's model catalog and understand how it differs from calling providers directly.
2Learn when Bedrock's AWS-native integration (IAM, VPC) matters for an enterprise use case.
3Compare Bedrock pricing against calling a provider's API directly.