Data & AI Lab Book a discovery call

Home / Roadmaps / AWS AI Engineer

Interactive learning roadmap

Choose your
AWS AI path

AWS has three AI-focused certifications for different goals. Compare the routes first, choose the one that matches your work, then open the detailed learning topics.

Choose a pathExplore all topics

Start with the destination

One foundation. Three separate AWS AI routes.

The AI Practitioner and Business Strategist paths are foundational but serve different roles. The Generative AI Developer path is an advanced builder route.

STARTWhat work do you want to do with AI?Choose the closest role below
01

For people who use, explain or support AI solutions

Foundational · Start here

AWS Certified AI Practitioner

  1. AI and ML concepts
  2. Generative AI fundamentals
  3. AWS AI services
  4. Responsible AI and security
  5. Exam preparation
Open this roadmap ↓Official exam page ↗
02

For experienced developers building production GenAI systems

Professional · Advanced path

AWS Certified Generative AI Developer – Professional

  1. Cloud and development foundations
  2. Amazon Bedrock and model choice
  3. RAG, agents and guardrails
  4. Deployment and integration
  5. Evaluation, security and operations
Open this roadmap ↓Official exam page ↗
03

For leaders making AI strategy, value and governance decisions

Foundational · Business path

AWS Certified AI Business Strategist

  1. AI business fundamentals
  2. Use-case discovery and value
  3. Responsible AI and governance
  4. Adoption and operating model
  5. Exam preparation
Open this roadmap ↓Official exam page ↗

Three independent roadmaps

Study only the path that matches your goal.

Similar concepts appear where each certification needs them, but their emphasis and learning outcomes differ.

For people who use, explain or support AI solutions

AWS Certified AI Practitioner

Foundational · 5 steps

AWS Certified AI Practitioner

AI and ML fundamentals

Understand AI, ML, deep learning, inference and common business use cases.

What you should be able to do

  • Explain the difference between AI, ML and deep learning
  • Recognize prediction, classification and generative use cases

Recommended resources

AWS AI services overviewaws.amazon.com ↗AWS Skill Builderskillbuilder.aws ↗

AWS Certified AI Practitioner

Generative AI foundations

Learn foundation models, tokens, prompts, embeddings, limitations and responsible use.

What you should be able to do

  • Describe how foundation models generate output
  • Recognize hallucination, bias and prompt risks

Recommended resources

Generative AI on AWSaws.amazon.com ↗Amazon Bedrock overviewaws.amazon.com ↗

AWS Certified AI Practitioner

AWS AI services

Match AWS managed AI services and Amazon Bedrock capabilities to common needs.

What you should be able to do

  • Choose between pre-trained AI services and custom ML
  • Identify suitable AWS services for text, speech, vision and search

Recommended resources

AWS AI servicesaws.amazon.com ↗Amazon Bedrock documentationdocs.aws.amazon.com ↗

AWS Certified AI Practitioner

Responsible and secure AI

Understand fairness, explainability, privacy, security and governance controls.

What you should be able to do

  • Identify responsible AI dimensions
  • Apply basic data protection and access principles

Recommended resources

AWS Responsible AIaws.amazon.com ↗AWS AI securityaws.amazon.com ↗

AWS Certified AI Practitioner

Practitioner exam readiness

Review the official exam guide, close knowledge gaps and practise scenario questions.

What you should be able to do

  • Map knowledge to the current exam domains
  • Build a focused final revision plan

Recommended resources

Official certification pageaws.amazon.com ↗AWS certification preparationaws.amazon.com ↗
Foundation Core Advanced Milestone

Interactive curriculum canvas

Developer deep curriculum: follow each stage.

Select any topic card to open its explanation, learning outcomes and recommended resources. Mark it complete when you are comfortable applying it.

Foundation topic

AWS cloud basics

Learn Regions, Availability Zones, shared responsibility, service categories and the AWS pricing model.

What you should be able to do

  • Explain the AWS global infrastructure
  • Separate AWS responsibilities from customer responsibilities
  • Estimate and monitor basic service costs

Primary learning resources

AWS Cloud Essentialsaws.amazon.com ↗AWS Cloud Practitioner Essentialsaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
IAM and least privilege AWS service icon

Foundation topic

IAM and least privilege

Identity is the control plane for every AI workload. Learn roles, policies, temporary credentials and permission boundaries.

What you should be able to do

  • Prefer roles and short-lived credentials
  • Read and write identity policies
  • Separate human, workload and deployment access

Primary learning resources

IAM documentationdocs.aws.amazon.com ↗IAM best practicesdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Foundation topic

Python, APIs and Git

Use Python to call services, transform data and automate deployments. Understand HTTP, JSON, SDK clients and version control.

What you should be able to do

  • Write testable Python modules
  • Call AWS APIs with Boto3
  • Use branches, pull requests and environment variables

Primary learning resources

AWS SDK for Pythondocs.aws.amazon.com ↗Git and GitHub roadmaproadmap.sh ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Foundation topic

Networking, CLI and IaC

Learn VPC concepts, private connectivity, the AWS CLI and infrastructure as code so environments are reproducible.

What you should be able to do

  • Recognize subnets, routes and security groups
  • Use profiles and the AWS CLI safely
  • Deploy a small stack with CloudFormation or CDK

Primary learning resources

VPC documentationdocs.aws.amazon.com ↗AWS CDK guidedocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Foundation topic

Machine learning lifecycle

Move from a measurable problem to data preparation, training, evaluation, deployment and monitoring.

What you should be able to do

  • Choose a useful target and baseline
  • Split data without leakage
  • Distinguish offline evaluation from production impact

Primary learning resources

AWS machine learning learning planexplore.skillbuilder.aws ↗Machine learning on AWSaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Foundation topic

Core ML concepts

Learn supervised and unsupervised learning, features, labels, overfitting, common metrics and model-selection trade-offs.

What you should be able to do

  • Select metrics for classification and regression
  • Explain bias and variance
  • Recognize when a rules-based system is enough

Primary learning resources

ML terminologydocs.aws.amazon.com ↗AWS AI and ML videoswww.youtube.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Foundation topic

Generative AI basics

Understand foundation models, tokens, context windows, embeddings, prompt patterns, retrieval and agents.

What you should be able to do

  • Compare prompting, RAG and fine-tuning
  • Explain embeddings and semantic search
  • Identify hallucination and prompt-injection risks

Primary learning resources

Generative AI on AWSaws.amazon.com ↗AWS generative AI learning planexplore.skillbuilder.aws ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Foundation topic

Responsible AI

Design for fairness, explainability, privacy, safety, transparency and human oversight throughout the lifecycle.

What you should be able to do

  • Document intended use and limitations
  • Evaluate harmful or biased outcomes
  • Create escalation and review paths

Primary learning resources

AWS responsible AIaws.amazon.com ↗Responsible AI resourcesaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Amazon S3 data lake AWS service icon

Core topic

Amazon S3 data lake

Use S3 as durable object storage for raw, curated, training and evaluation datasets with deliberate layout and lifecycle rules.

What you should be able to do

  • Design prefixes and formats for analytical access
  • Use versioning, encryption and lifecycle policies
  • Control access to sensitive training data

Primary learning resources

Amazon S3 documentationdocs.aws.amazon.com ↗Data lake on AWSaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
AWS Glue and data catalog AWS service icon

Core topic

AWS Glue and data catalog

Discover schemas, maintain a catalog and build managed ETL jobs for repeatable data preparation.

What you should be able to do

  • Catalog datasets and partitions
  • Build and monitor ETL jobs
  • Track schema and data-quality changes

Primary learning resources

AWS Glue documentationdocs.aws.amazon.com ↗AWS Glue Data Qualitydocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Streaming and events AWS service icon

Core topic

Streaming and events

Use streams and events when AI applications need fresh context, asynchronous workflows or change-driven processing.

What you should be able to do

  • Distinguish streams, queues and event buses
  • Design idempotent consumers
  • Handle retries, ordering and dead-letter paths

Primary learning resources

Kinesis Data Streamsdocs.aws.amazon.com ↗Amazon EventBridgedocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Core topic

Data governance and quality

Define ownership, classification, lineage, retention and quality checks before models consume the data.

What you should be able to do

  • Create measurable data contracts
  • Protect personal and confidential data
  • Record lineage and model-dataset versions

Primary learning resources

AWS data governanceaws.amazon.com ↗AWS Well-Architected ML Lensdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Amazon SageMaker AI AWS service icon

Core topic

Amazon SageMaker AI

Build, train, tune and deploy custom machine-learning models using managed development and inference capabilities.

What you should be able to do

  • Use Studio for experiments
  • Run managed training and tuning jobs
  • Choose real-time, serverless, batch or asynchronous inference

Primary learning resources

SageMaker AI documentationdocs.aws.amazon.com ↗SageMaker AI workshopscatalog.workshops.aws ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Amazon Bedrock and model choice AWS service icon

Core topic

Amazon Bedrock and model choice

Build generative AI applications with managed foundation models while comparing quality, latency, context and cost.

What you should be able to do

  • Invoke models through the Bedrock APIs
  • Evaluate models with representative tasks
  • Manage inference settings and cost

Primary learning resources

Amazon Bedrock documentationdocs.aws.amazon.com ↗Amazon Bedrock workshopcatalog.workshops.aws ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Core topic

Prompting and evaluation

Treat prompts and model settings as versioned application components with datasets, rubrics and regression tests.

What you should be able to do

  • Write structured prompts and output contracts
  • Build task-specific evaluation sets
  • Measure quality, safety, latency and cost together

Primary learning resources

Bedrock model evaluationdocs.aws.amazon.com ↗Prompt engineering guidelinesdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
RAG and semantic search AWS service icon

Core topic

Ground model responses in approved enterprise content using chunking, embeddings, retrieval and source attribution.

What you should be able to do

  • Design ingestion and chunking
  • Tune retrieval independently of generation
  • Evaluate groundedness and citation quality

Primary learning resources

Knowledge Bases for Bedrockdocs.aws.amazon.com ↗Amazon OpenSearch vector searchdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Agents, tools and guardrails AWS service icon

Advanced topic

Agents, tools and guardrails

Build multi-step systems that call tools safely, maintain state and apply policy controls around inputs and outputs.

What you should be able to do

  • Define narrow tool contracts and permissions
  • Control prompt injection and unsafe actions
  • Trace agent decisions and failures

Primary learning resources

Agents for Amazon Bedrockdocs.aws.amazon.com ↗Guardrails for Amazon Bedrockdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Lambda and API Gateway AWS service icon

Core topic

Lambda and API Gateway

Expose model workflows through secure, observable APIs while keeping synchronous paths small and predictable.

What you should be able to do

  • Validate and limit requests
  • Use timeouts, retries and idempotency
  • Stream or queue work that exceeds request limits

Primary learning resources

AWS Lambda documentationdocs.aws.amazon.com ↗API Gateway documentationdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Step Functions orchestration AWS service icon

Core topic

Step Functions orchestration

Coordinate ingestion, evaluation, human review and long-running AI workflows with explicit state and failure handling.

What you should be able to do

  • Model workflows as states
  • Add retries, catches and compensating actions
  • Keep execution inputs free of unnecessary sensitive data

Primary learning resources

AWS Step Functionsdocs.aws.amazon.com ↗Bedrock and Step Functionsdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Event-driven AI AWS service icon

Advanced topic

Event-driven AI

React to business events without tightly coupling producers to model services or downstream actions.

What you should be able to do

  • Route events with stable contracts
  • Protect consumers from duplicate delivery
  • Observe end-to-end business outcomes

Primary learning resources

Amazon EventBridge documentationdocs.aws.amazon.com ↗Event-driven architectureaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Advanced topic

Human review and feedback

Design review queues and feedback loops for uncertain, high-impact or policy-sensitive model decisions.

What you should be able to do

  • Define confidence and escalation thresholds
  • Capture reviewer decisions consistently
  • Use feedback without leaking restricted data

Primary learning resources

Amazon Augmented AIdocs.aws.amazon.com ↗SageMaker Ground Truthdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Security and encryption AWS service icon

Core topic

Security and encryption

Use least privilege, KMS encryption, private networking and data boundaries appropriate to the workload.

What you should be able to do

  • Threat-model data and model access
  • Use customer-managed keys when required
  • Keep secrets out of code and logs

Primary learning resources

AWS KMS documentationdocs.aws.amazon.com ↗AWS security best practicesaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
Observability and model monitoring AWS service icon

Core topic

Observability and model monitoring

Monitor availability, latency, cost, data quality and model behavior instead of relying on infrastructure metrics alone.

What you should be able to do

  • Create CloudWatch metrics and alarms
  • Trace model and tool calls safely
  • Detect drift and quality regression

Primary learning resources

Amazon CloudWatchdocs.aws.amazon.com ↗SageMaker Model Monitordocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.
MLOps and delivery AWS service icon

Advanced topic

MLOps and delivery

Version code, data, prompts and models; automate tests and promote changes through controlled environments.

What you should be able to do

  • Build repeatable pipelines
  • Use model and prompt evaluation gates
  • Plan rollback and audit evidence

Primary learning resources

SageMaker MLOpsdocs.aws.amazon.com ↗AWS MLOps guidanceaws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Advanced topic

Cost and reliability

Balance model quality with token usage, compute, storage, availability targets and graceful degradation.

What you should be able to do

  • Estimate unit economics per workflow
  • Set budgets, quotas and rate limits
  • Design fallbacks for model and dependency failures

Primary learning resources

AWS Cost Managementdocs.aws.amazon.com ↗Generative AI Lensdocs.aws.amazon.com ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Milestone topic

Production portfolio project

Ship one end-to-end solution with infrastructure as code, evaluation evidence, security controls, monitoring and a concise architecture decision record.

What you should be able to do

  • Demonstrate a real business workflow
  • Publish architecture and trade-offs without exposing secrets
  • Measure quality, latency and estimated cost

Primary learning resources

AWS Samplesgithub.com ↗AWS Workshopsworkshops.aws ↗

Links open official or clearly identified external resources. Verify current AWS service and certification details before relying on them.

Editorial method

An independent learning guide.

This roadmap was created by Data & AI Lab from practical engineering experience and current AWS primary documentation. It is not copied from roadmap.sh and is not an official AWS curriculum.

AWS, Amazon Bedrock, SageMaker and other AWS service names and icons are trademarks or assets of Amazon Web Services, Inc. Their use identifies the services discussed and does not imply endorsement or partnership.