AWS Certified Generative AI Developer - Professional (AIP-C01) Exam Prep

What's inside

1. Level 1: How AIP-C01 Works and How to Use This CourseFree
   [Exam Orientation] Covers what AWS's AIP-C01 exam guide says about the exam: 65 scored and 10 unscored multiple-choice and multiple-response questions, a pass mark of 750 on a 100–1,000 scaled score, compensatory scoring and the five domain weightings. It also says where to get the guide free and that this course is independent of AWS.
2. Level 2: Designing GenAI Architectures from RequirementsFree
   [Domain 1: Foundation Model Integration, Data Management, and Compliance — Task 1.1, Skill 1.1.1] Turns business needs and technical constraints into a GenAI architecture by choosing suitable FMs, integration patterns and deployment strategies.
3. Level 3: Proofs of Concept and the Well-Architected GenAI Lens
   [Domain 1 — Task 1.1, Skills 1.1.2–1.1.3] Builds a proof of concept on Amazon Bedrock to test feasibility, performance and business value, then standardises components with the AWS Well-Architected Framework and the AWS WA Tool Generative AI Lens.
4. Level 4: Assessing and Choosing Foundation Models
   [Domain 1 — Task 1.2, Skill 1.2.1] Evaluates and picks FMs on Amazon Bedrock using performance benchmarks, capability analysis and limitation evaluation against the use case.
5. Level 5: Dynamic Model Selection Without Code Changes
   [Domain 1 — Task 1.2, Skill 1.2.2] Builds flexible architectures that switch models or providers at runtime using AWS Lambda, Amazon API Gateway and AWS AppConfig.
6. Level 6: Resilient FM Systems, Cross-Region Inference, Residency
   [Domain 1 — Task 1.2, Skill 1.2.3] Keeps AI systems running during service disruptions with AWS Step Functions circuit breaker patterns, Amazon Bedrock Cross-Region Inference, cross-Region model deployment and graceful degradation, including the data-residency trade-offs that inference profiles bring.
7. Level 7: FM Customisation Deployment and Lifecycle
   [Domain 1 — Task 1.2, Skill 1.2.4] Manages fine-tuned FMs from deployment to retirement: deploying domain-specific models on Amazon SageMaker AI, LoRA and adapters, versioning in SageMaker Model Registry, automated pipelines and rollback.
8. Level 8: Data Validation Workflows for FM Consumption
   [Domain 1 — Task 1.3, Skill 1.3.1] Makes sure data meets quality standards before FMs use it, with AWS Glue Data Quality, SageMaker Data Wrangler, custom Lambda functions and Amazon CloudWatch metrics.
9. Level 9: Processing Multimodal Data: Text, Image, Audio, Tables
   [Domain 1 — Task 1.3, Skill 1.3.2] Builds processing workflows for text, image, audio and tabular data using Amazon Bedrock multimodal models, SageMaker Processing and Amazon Transcribe.
10. Level 10: Formatting and Enhancing Input Data for FMs
    [Domain 1 — Task 1.3, Skills 1.3.3–1.3.4] Formats inference input to each model's requirements and improves input quality with Amazon Bedrock reformatting, Amazon Comprehend entity extraction and Lambda normalisation.
11. Level 11: Vector Store Architectures and Metadata Frameworks
    [Domain 1 — Task 1.4, Skills 1.4.1–1.4.2] Designs vector database architectures for FM augmentation: Amazon Bedrock Knowledge Bases, Amazon OpenSearch Service with the Neural plugin, Amazon RDS with Amazon S3 document repositories, and Amazon DynamoDB holding metadata alongside embeddings, plus metadata frameworks that sharpen retrieval.
12. Level 12: High-Performance Vector Search at Scale
    [Domain 1 — Task 1.4, Skill 1.4.3] Tunes semantic search for scale using OpenSearch sharding strategies, multi-index approaches for specialised domains and hierarchical indexing techniques.
13. Level 13: Data Source Integration and Vector Store Freshness
    [Domain 1 — Task 1.4, Skills 1.4.4–1.4.5] Connects document management systems, knowledge bases and internal wikis to GenAI applications, and keeps vector stores current with incremental updates, change detection, automated synchronisation and scheduled refreshes.
14. Level 14: Chunking Strategies and Choosing Embedding Models
    [Domain 1 — Task 1.5, Skills 1.5.1–1.5.2] Splits documents with Amazon Bedrock chunking options, Lambda fixed-size chunking and custom hierarchical chunking, and chooses embedding models such as Amazon Titan embeddings by dimensionality and domain fit.
15. Level 15: Vector Search Deployment, Hybrid Search and Reranking
    [Domain 1 — Task 1.5, Skills 1.5.3–1.5.4] Deploys semantic search on OpenSearch Service, Amazon Aurora with pgvector or Bedrock Knowledge Bases' managed vector store, then improves relevance with hybrid keyword-plus-vector search and Amazon Bedrock reranker models.
16. Level 16: Query Transformation and Standard Retrieval Access
    [Domain 1 — Task 1.5, Skills 1.5.5–1.5.6] Improves retrieval with query expansion in Amazon Bedrock, query decomposition in Lambda and query transformation in Step Functions, and exposes retrieval to FMs through function calling, MCP clients and standard API patterns.
17. Level 17: Instruction Frameworks and Iterative Prompt Refinement
    [Domain 1 — Task 1.6, Skills 1.6.1 and 1.6.5] Controls FM behaviour with Amazon Bedrock Prompt Management role definitions, Amazon Bedrock Guardrails and response templates, then refines prompts with structured input, output format specifications, chain-of-thought instructions and feedback loops.
18. Level 18: Prompt Governance and Quality Assurance
    [Domain 1 — Task 1.6, Skills 1.6.3–1.6.4] Governs prompts with Amazon Bedrock Prompt Management parameterised templates and approval workflows, Amazon S3 template repositories, AWS CloudTrail usage tracking and CloudWatch Logs, and tests prompt quality with Lambda output checks, Step Functions edge-case tests and CloudWatch regression monitoring.
19. Level 19: Conversational Context and Prompt Flows
    [Domain 1 — Task 1.6, Skills 1.6.2 and 1.6.6] Builds interactive systems that keep context, using DynamoDB conversation history, Amazon Comprehend intent recognition and Step Functions clarification workflows, and designs complex prompt systems with Amazon Bedrock Prompt Flows.
20. Level 20: Agents with Memory and State: Strands and MCP
    [Domain 2: Implementation and Integration — Task 2.1, Skill 2.1.1] Builds autonomous agents with suitable memory and state management using Strands Agents, uses AWS Agent Squad for multi-agent systems, and uses MCP for agent-tool interactions.
21. Level 21: Agent Memory and State with Strands and AgentCore
    [Domain 2: Implementation and Integration — Task 2.1, Skill 2.1.1] Learn how agents keep track of a conversation and keep state between steps. Build a Strands Agents agent in boto3-backed Python, then compare short-term session state, long-term memory and where Amazon Bedrock AgentCore Memory fits when the agent runs in production.
22. Level 22: Multi-Agent Systems and Model Coordination
    [Domain 2: Implementation and Integration — Task 2.1, Skills 2.1.1 and 2.1.4] Design multi-agent systems where a supervisor or classifier sends work to specialist agents, using AWS Agent Squad and Strands Agents. Coordinate several specialised FMs with model selection frameworks and custom aggregation logic for model ensembles.
23. Level 23: MCP Servers: Lambda vs Amazon ECS
    [Domain 2: Implementation and Integration — Task 2.1, Skills 2.1.1 and 2.1.7] Extend FMs with the Model Context Protocol (MCP). Build stateless MCP servers on Lambda for lightweight tool access, decide when complex or long-running tools need MCP servers on Amazon ECS, and use MCP client libraries so every agent reaches tools the same way.
24. Level 24: Reliable Tool Integrations and Function Definitions
    [Domain 2: Implementation and Integration — Task 2.1, Skill 2.1.6] Make tool calls reliable. Write standardised function definitions (JSON Schema tool specs), add custom behaviour through the Strands API, and put parameter validation and error handling in the Lambda functions that carry out the tools.
25. Level 25: Structured Reasoning: ReAct and Chain-of-Thought
    [Domain 2: Implementation and Integration — Task 2.1, Skill 2.1.2] Give FMs a way to break down complex problems. Use AWS Step Functions to orchestrate a ReAct (reason, act, observe) loop and prompt for chain-of-thought reasoning, so each step can be seen and retried.
26. Level 26: Safeguarded Agent Workflows and Human Review
    [Domain 2: Implementation and Integration — Task 2.1, Skills 2.1.3 and 2.1.5] Keep agent behaviour under control with Step Functions stopping conditions, Lambda timeouts, IAM policies that limit what resources an agent can touch, and circuit breakers. Add human augmentation with Step Functions review and approval steps and API Gateway feedback collection.
27. Level 27: FM Deployment Strategies: Bedrock and SageMaker AI
    [Domain 2: Implementation and Integration — Task 2.2, Skills 2.2.1-2.2.3] Choose how to deploy FMs for a workload: on-demand Bedrock invocation from Lambda, Amazon Bedrock provisioned throughput, or SageMaker AI endpoints for hybrid solutions. Cover what makes LLMs different to deploy (memory, GPU use, token capacity, model loading), when to use smaller task-specific models, and API-based model cascading.
28. Level 28: Enterprise Connectivity and Event-Driven Integration
    [Domain 2: Implementation and Integration — Task 2.3, Skills 2.3.1 and 2.3.2] Bring FM capabilities into existing enterprise systems. Use API-based integrations with legacy systems, data synchronisation patterns, API Gateway microservice integrations, Lambda webhook handlers and Amazon EventBridge for loosely coupled, event-driven designs.
29. Level 29: Secure Access and Cross-Environment AI
    [Domain 2: Implementation and Integration — Task 2.3, Skills 2.3.3 and 2.3.4] Build secure access frameworks with identity federation between FM services and enterprise identity, role-based access control for models and data, and least-privilege API access to FMs. Meet data rules across jurisdictions with AWS Outposts, AWS Wavelength and secure routing between cloud and on-premises resources.
30. Level 30: CI/CD Pipelines and GenAI Gateway Architectures
    [Domain 2: Implementation and Integration — Task 2.3, Skill 2.3.5] Ship GenAI components safely with AWS CodePipeline and AWS CodeBuild, including automated tests, security scans and rollback. Put a GenAI gateway in front of all FM use: a central abstraction layer with observability and controls.
31. Level 31: FM API Integration: Sync, Async and Streaming
    [Domain 2: Implementation and Integration — Tasks 2.4 and 2.5, Skills 2.4.1, 2.4.2 and 2.5.1] Build flexible model interaction systems: synchronous Bedrock calls (Converse, InvokeModel), asynchronous processing with Amazon SQS, and API Gateway request validation. Add real-time responses with Bedrock streaming APIs (ConverseStream), WebSockets or server-sent events, token limit management and retries for model timeouts.
32. Level 32: Resilient FM Systems and Intelligent Model Routing
    [Domain 2: Implementation and Integration — Task 2.4, Skills 2.4.3 and 2.4.4] Make FM systems reliable with AWS SDK exponential backoff, API Gateway rate limiting, fallback for graceful degradation and AWS X-Ray tracing across services. Route requests to the right model with static configuration, Step Functions content-based routing, routing based on metrics, and API Gateway request transformations.
33. Level 33: Accessible AI Interfaces and Business Enhancements
    [Domain 2: Implementation and Integration — Task 2.5, Skills 2.5.2, 2.5.3 and 2.5.5] Speed up adoption with AWS Amplify UI components, OpenAPI-first API design and Amazon Bedrock Prompt Flows for no-code workflows. Improve business systems with Lambda CRM enhancements, Step Functions document processing, Amazon Bedrock Data Automation, prompt chaining, and agent design patterns orchestrated with Strands Agents, AWS Agent Squad and Step Functions.
34. Level 34: Developer Productivity and FM Troubleshooting Tools
    [Domain 2: Implementation and Integration — Task 2.5, Skills 2.5.4 and 2.5.6] Use Amazon Q Developer to generate, refactor and test GenAI code. Troubleshoot FM applications faster by querying prompts and responses with CloudWatch Logs Insights, tracing FM API calls with X-Ray and spotting GenAI-specific error patterns.
35. Level 35: Amazon Bedrock Guardrails for Harmful Inputs
    [Domain 3: AI Safety, Security, and Governance — Task 3.1, Skill 3.1.1] Protect FMs from harmful user input. Set up Amazon Bedrock guardrails (content filters, denied topics, word filters), apply them with the ApplyGuardrail API, and build custom moderation workflows and real-time validation with Step Functions and Lambda.
36. Level 36: Output Safety and Hallucination Reduction
    [Domain 3: AI Safety, Security, and Governance — Task 3.1, Skills 3.1.2 and 3.1.3] Stop harmful or inaccurate outputs. Filter responses with Amazon Bedrock guardrails, use specialised FM evaluations for toxicity, and use text-to-SQL for deterministic answers. Ground responses with an Amazon Bedrock Knowledge Base, then verify them with confidence scoring, semantic similarity checks and JSON Schema for structured outputs.
37. Level 37: Defense-in-Depth Safety Architectures
    [Domain 3: AI Safety, Security, and Governance — Task 3.1, Skill 3.1.4] Layer protections against FM misuse: Amazon Comprehend pre-processing filters, model-based guardrails in Amazon Bedrock, Lambda post-processing validation and API Gateway response filtering.
38. Level 38: Prompt Injection and Jailbreak Defense
    [Domain 3: AI Safety, Security, and Governance — Task 3.1, Skill 3.1.5] Detect adversarial input and close security gaps. Combine prompt injection and jailbreak detection, input sanitisation and content filters, and safety classifiers, and set up automated adversarial testing workflows.
39. Level 39: Protected AI Environments: VPC, IAM, Lake Formation
    [Domain 3: AI Safety, Security, and Governance — Task 3.2, Skill 3.2.1] Secure FM deployments end to end: VPC endpoints (AWS PrivateLink) to keep network traffic private, IAM policies for secure data access, AWS Lake Formation for fine-grained data access, and CloudWatch to monitor data access.
40. Level 40: PII Protection and Privacy-Preserving FM Use
    [Domain 3: AI Safety, Security, and Governance — Task 3.2, Skills 3.2.2 and 3.2.3] Protect sensitive information in FM interactions. Detect PII with Amazon Comprehend and Amazon Macie, rely on Amazon Bedrock's built-in data privacy features, filter outputs with guardrail sensitive information filters, apply masking and anonymisation, and enforce retention with Amazon S3 Lifecycle configurations.
41. Level 41: Privacy-Preserving FM Interactions and PII Handling
    [Domain 3: AI Safety, Security, and Governance – Task 3.2, Skills 3.2.2 and 3.2.3] Learn to protect sensitive data before, during and after FM calls by finding PII with Amazon Comprehend and Amazon Macie, masking and anonymising it, filtering outputs with Amazon Bedrock guardrails, using Amazon Bedrock's built-in data privacy features, and enforcing retention with Amazon S3 Lifecycle configurations.
42. Level 42: Compliance Frameworks: Model Cards, Lineage, Decision Logs
    [Domain 3: AI Safety, Security, and Governance – Task 3.3, Skill 3.3.1] Learn to build a regulatory compliance framework for FM deployments using programmatic SageMaker AI model cards, automatic data lineage tracking with AWS Glue, metadata tagging for data source attribution, and full decision logs in CloudWatch Logs.
43. Level 43: Data Source Tracking and Audit Traceability
    [Domain 3: AI Safety, Security, and Governance – Task 3.3, Skill 3.3.2] Learn to make every FM answer traceable to its data: register sources in the AWS Glue Data Catalog, carry metadata tags through to the content the FM generates, and use CloudTrail for audit logging of who called which model and resource.
44. Level 44: Organisational Governance and Continuous Oversight
    [Domain 3: AI Safety, Security, and Governance – Task 3.3, Skills 3.3.3 and 3.3.4] Learn to set up organisational governance for FM implementations that follows policy, regulation and responsible AI principles, then keep it running with automated detection of misuse, drift, bias drift and policy violations, alerting and remediation workflows, token-level redaction, response logging and AI output policy filters.
45. Level 45: Transparent AI: Reasoning, Confidence and Evidence
    [Domain 3: AI Safety, Security, and Governance – Task 3.4, Skill 3.4.1] Learn to make FM outputs transparent to users and auditors with reasoning displays, confidence metrics and uncertainty measures collected in CloudWatch, evidence presentation for source attribution, and Amazon Bedrock agent tracing for reasoning traces.
46. Level 46: Fairness Evaluation and Policy-Compliant AI
    [Domain 3: AI Safety, Security, and Governance – Task 3.4, Skills 3.4.2 and 3.4.3] Learn to check FM outputs for bias with pre-defined fairness metrics in CloudWatch, systematic A/B tests through Amazon Bedrock Prompt Management and Prompt Flows, and LLM-as-a-judge evaluations, then enforce responsible AI policy with guardrails, model cards that document limitations, and Lambda compliance checks.
47. Level 47: Token Efficiency: Estimation, Pruning, Response Limits
    [Domain 4: Operational Efficiency and Optimization – Task 4.1, Skill 4.1.1] Learn to cut FM spend without losing quality through token estimation and tracking, context window optimisation, prompt compression, context pruning, and response size controls and limits.
48. Level 48: Cost-Effective Model Selection and Tiered FM Usage
    [Domain 4: Operational Efficiency and Optimization – Task 4.1, Skill 4.1.2] Learn to build model selection frameworks that weigh cost against capability, send queries to different FM tiers depending on how complex they are, balance inference cost against response quality, and measure price-to-performance.
49. Level 49: Throughput, Capacity Planning and Provisioned Throughput
    [Domain 4: Operational Efficiency and Optimization – Task 4.1, Skill 4.1.3; Task 4.2, Skills 4.2.3 and 4.2.5] Learn to get the most utilisation and throughput from FM workloads with batching strategies, batch inference, management of concurrent model invocations, capacity planning for token processing, utilisation monitoring of prompt and completion patterns, auto-scaling tuned for GenAI traffic, and provisioned throughput optimisation.
50. Level 50: Intelligent Caching: Prompt, Semantic and Edge
    [Domain 4: Operational Efficiency and Optimization – Task 4.1, Skill 4.1.4] Learn to avoid unnecessary FM invocations with prompt caching, semantic caching, result fingerprinting, deterministic request hashing and edge caching, and to know when each layer pays off.
51. Level 51: Responsive AI: Latency, Streaming and Parameter Tuning
    [Domain 4: Operational Efficiency and Optimization – Task 4.2, Skills 4.2.1, 4.2.4 and 4.2.6] Learn to manage latency-cost trade-offs and tune FM performance with pre-computation, latency-optimised Amazon Bedrock models, parallel requests, response streaming, benchmarking, API call profiling, model-specific temperature and top-k/top-p settings, and A/B testing.
52. Level 52: Retrieval Performance and Vector Store Operations
    [Domain 4: Operational Efficiency and Optimization – Task 4.2, Skill 4.2.2; Task 4.3, Skill 4.3.5] Learn to make retrieval faster and more relevant through index optimisation, query preprocessing and hybrid search with custom scoring, and to run vector stores reliably with performance monitoring, automated index optimisation and data quality validation.
53. Level 53: GenAI Observability and Model Invocation Logs
    [Domain 4: Operational Efficiency and Optimization – Task 4.3, Skills 4.3.1–4.3.4 and 4.3.6] Learn to build full observability for FM applications: operational and business metrics on CloudWatch dashboards, FM interaction tracing, Amazon Bedrock Model Invocation Logs, anomaly detection for token bursts, response drift and cost, tool-calling and multi-agent tracking, and GenAI-specific failure detection with golden datasets, output diffing and reasoning path tracing.
54. Level 54: FM Assessment Frameworks and Bedrock Model Evaluations
    [Domain 5: Testing, Validation, and Troubleshooting – Task 5.1, Skills 5.1.1 and 5.1.2] Learn to judge FM output quality beyond traditional ML metrics (relevance, factual accuracy, consistency and fluency) and to find the best configuration with Amazon Bedrock Model Evaluations, multi-model comparison, A/B and canary testing, and cost-performance analysis.
55. Level 55: LLM-as-a-Judge, RAG Evaluation and Human Feedback
    [Domain 5: Testing, Validation, and Troubleshooting – Task 5.1, Skills 5.1.3, 5.1.5 and 5.1.6] Learn to evaluate from several angles using LLM-as-a-Judge, RAG evaluation of knowledge bases, retrieval quality testing (relevance scoring, context matching, retrieval latency), and user-centred feedback through rating systems and annotation workflows.
56. Level 56: Agent Evaluation and Stakeholder Reporting
    [Domain 5: Testing, Validation, and Troubleshooting – Task 5.1, Skills 5.1.7 and 5.1.8] Learn to check that agents complete tasks correctly and efficiently by measuring task completion rate, tool usage effectiveness and multi-step reasoning quality with Amazon Bedrock Agent evaluations, then report the results to stakeholders with visualisations, automated reports and model comparisons.
57. Level 57: Continuous QA, Regression Tests and Deployment Gates
    [Domain 5: Testing, Validation, and Troubleshooting – Task 5.1, Skills 5.1.4 and 5.1.9] Learn to keep FM quality steady through updates with continuous evaluation workflows, regression testing of model outputs, automated quality gates in deployment pipelines, synthetic user workflows, and validation of hallucination rate, semantic drift and response consistency.
58. Level 58: Troubleshooting Content Handling and FM API Errors
    [Domain 5: Testing, Validation, and Troubleshooting – Task 5.2, Skills 5.2.1 and 5.2.2] Learn to diagnose and fix incomplete processing and integration failures: context window overflow, truncation errors, dynamic chunking fixes, prompt design changes, and Bedrock API problems found through error logging, request validation and response analysis.
59. Level 59: Troubleshooting Prompts and Retrieval Systems
    [Domain 5: Testing, Validation, and Troubleshooting – Task 5.2, Skills 5.2.3, 5.2.4 and 5.2.5] Learn to fix prompt and retrieval quality problems systematically with prompt testing frameworks, version comparison, CloudWatch Logs and X-Ray prompt observability, schema validation for format inconsistencies, embedding quality diagnostics, drift monitoring, vectorisation and chunking fixes, and vector search tuning.
60. Level 60: Timed Mock Exam: AIP-C01 (65 Questions, 170 Minutes) Timed mock
    [All Domains] A timed mock exam of 65 original scenario questions, multiple choice and multiple response, weighted across all five domains like the exam guide (31/26/20/12/11%), with a 75% pass mark and every distractor explained.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →