CloudIntermediatev2026Updated 2026-07

Amazon Web Services

Q: What is Amazon Bedrock and how does it differ from SageMaker?

Amazon Bedrock provides serverless API access to third-party foundation models (Claude, Llama, Mistral) and Amazon's own Titan models. You pay per token with no infrastructure to manage. SageMaker is a comprehensive ML platform for training custom models, running notebooks, and deploying your own model endpoints. Use Bedrock for leveraging existing FMs; use SageMaker for building and fine-tuning custom models.

Q: What is the AWS free tier and what does it include?

The AWS Free Tier offers always-free, 12-month free, and trial services. Always-free includes Lambda (1M requests/month), DynamoDB (25GB), S3 (5GB). 12-month free includes EC2 t2.micro (750 hours/month), RDS db.t2.micro (750 hours/month), and CloudFront (1TB transfer). New accounts also get $300 credit for some services.

Q: Lambda vs ECS Fargate — when to use each?

Use Lambda for event-driven workloads with variable traffic, short execution times (<15 min), and minimal runtime dependencies. Use ECS Fargate for long-running containers, apps needing persistent connections, workloads that require specific runtime control, or services that need consistent performance without cold starts.

The world's most comprehensive cloud platform

CloudServerlessS3LambdaBedrockIaC

Overview

1AWS maintains 33% global cloud market share in 2026 with 200+ services across 33 regions.
2Amazon Bedrock is the fastest-growing AWS service — access Claude, Llama, Titan, and Mistral via unified API.
3Amazon Q Developer (formerly CodeWhisperer) integrates AI pair programming in IDE and CLI.
4Aurora DSQL provides globally distributed serverless SQL with zero-infrastructure management.
5Graviton4 instances deliver the best price/performance for compute-intensive workloads.

Key Features in 2026

Amazon Bedrock — managed API for 50+ foundation models including Claude, Llama, Mistral

Amazon Q Developer — AI coding assistant in IDE, CLI, and AWS Console

Aurora DSQL — globally distributed, serverless, PostgreSQL-compatible SQL

S3 Express One Zone — 10x lower latency than S3 Standard for high-throughput workloads

Lambda SnapStart — pre-initialised execution environments (now supports Node.js, Python)

Graviton4 — AWS Arm-based compute with best price/performance in cloud

CloudFront KeyValueStore — edge key-value storage for Lambda@Edge functions

Use Cases

→Scalable web applications and APIs on Lambda + API Gateway + DynamoDB
→AI/ML workloads on SageMaker or Bedrock
→Data lakes and analytics with S3 + Glue + Athena + Redshift
→Container workloads on ECS Fargate or EKS

2026 Key Updates

Amazon Bedrock: 50+ FMs, Bedrock Agents with Code Interpreter, Knowledge Bases with RAG
Amazon Q Developer: in-IDE AI coding, /transform command for Java upgrades, Unit Test generation
Aurora DSQL GA: globally distributed SQL, 99.99% availability, zero management
S3 Express One Zone: single-digit ms latency, ideal for ML training data pipelines
Lambda SnapStart for Python and Node.js: pre-warming reduces cold starts to <1s
Graviton4 (r8g, c8g instances): 30% better price/performance over Graviton3
EC2 Trn2 instances with AWS Trainium2: cost-effective LLM training at scale

Core Services

EC2: compute instances from t4g.nano (free tier) to Trn2 96xlarge (AI training)
S3: 11 nines durability, intelligent tiering, lifecycle policies, event notifications
Lambda: serverless functions; 15-minute max timeout; 10GB memory; ARM/x86
RDS / Aurora: managed relational databases with Multi-AZ, read replicas, Proxy
DynamoDB: key-value/document DB; on-demand billing; Global Tables for multi-region
CloudFront: CDN with 600+ PoPs; integrated WAF, Shield, and Lambda@Edge

AI & ML Services

Amazon Bedrock: unified API for Claude (Anthropic), Llama (Meta), Mistral, Titan (AWS)
Bedrock Agents: multi-step AI agents with tool use and knowledge base retrieval
SageMaker: end-to-end ML platform for training, evaluation, and deployment
Amazon Rekognition: computer vision (object detection, face analysis)
Amazon Transcribe: speech-to-text with PII redaction and custom vocabulary
Amazon Q Business: generative AI for enterprise search and chat over internal data

Infrastructure as Code

CloudFormation: AWS-native IaC with YAML/JSON templates; StackSets for multi-account
AWS CDK: define infrastructure in TypeScript/Python/Java; compiles to CloudFormation
Terraform (HashiCorp): multi-cloud IaC; large community; state management in S3
AWS SAM: simplified CloudFormation for serverless apps (Lambda, API Gateway, DynamoDB)
AWS Config: drift detection and compliance rules for deployed resources
CI/CD: CodePipeline + CodeBuild + CodeDeploy or integrate with GitHub Actions

Frequently Asked Questions

What is Amazon Bedrock and how does it differ from SageMaker?

Amazon Bedrock provides serverless API access to third-party foundation models (Claude, Llama, Mistral) and Amazon's own Titan models. You pay per token with no infrastructure to manage. SageMaker is a comprehensive ML platform for training custom models, running notebooks, and deploying your own model endpoints. Use Bedrock for leveraging existing FMs; use SageMaker for building and fine-tuning custom models.

What is the AWS free tier and what does it include?

The AWS Free Tier offers always-free, 12-month free, and trial services. Always-free includes Lambda (1M requests/month), DynamoDB (25GB), S3 (5GB). 12-month free includes EC2 t2.micro (750 hours/month), RDS db.t2.micro (750 hours/month), and CloudFront (1TB transfer). New accounts also get $300 credit for some services.

Lambda vs ECS Fargate — when to use each?

Use Lambda for event-driven workloads with variable traffic, short execution times (<15 min), and minimal runtime dependencies. Use ECS Fargate for long-running containers, apps needing persistent connections, workloads that require specific runtime control, or services that need consistent performance without cold starts.