AIP-C01 Domain 1: Foundation Model Integration, Data Management, and Compliance (31%)
Domain 1 is the largest part of the AIP-C01 exam, worth 31%. It tests whether you can design a GenAI solution from requirements, choose and configure foundation models, prepare data for them, build vector stores and retrieval for RAG, and manage prompts with governance.
Weight on the exam: 31%. It is one of the five domains of the AWS Certified Generative AI Developer – Professional AIP-C01 exam. Below is a summary of its tasks and skills according to the official AWS guide.
Tasks and skills tested
Task 1.1: Analyze requirements and design GenAI solutions
- Create architectural designs that fit business needs and technical constraints, including the FMs, integration patterns and deployment strategies.
- Build proof-of-concept implementations (for example, with Amazon Bedrock) to validate feasibility, performance and business value before full deployment.
- Create standardized technical components for consistent implementation, using the AWS Well-Architected Framework and its Generative AI Lens.
Task 1.2: Select and configure FMs
- Assess and choose FMs against business use cases and technical requirements, using benchmarks, capability analysis and limitation evaluation.
- Create patterns for dynamic model selection and provider switching without code changes (for example, AWS Lambda, Amazon API Gateway, AWS AppConfig).
- Design resilient systems that keep running during disruptions: Step Functions circuit breakers, cross-Region inference, cross-Region deployment and graceful degradation.
- Manage deployment and lifecycle of customized FMs: SageMaker AI endpoints, parameter-efficient adaptation such as LoRA, Model Registry versioning, automated pipelines, rollback and model retirement.
Task 1.3: Implement data validation and processing pipelines for FM consumption
- Validate that data meets quality standards (AWS Glue Data Quality, SageMaker Data Wrangler, custom Lambda functions, CloudWatch metrics).
- Process text, image, audio and tabular data, including multimodal pipelines.
- Format input data to each model's requirements, such as JSON for Bedrock API requests and conversation formatting for dialog applications.
- Improve input quality to improve response quality and consistency, for example by reformatting text, extracting entities with Amazon Comprehend or normalizing data in Lambda.
Task 1.4: Design and implement vector store solutions
- Build vector database architectures for semantic retrieval: Bedrock Knowledge Bases, OpenSearch Service with the Neural plugin, RDS with S3 document repositories, DynamoDB for metadata and embeddings.
- Develop metadata frameworks (S3 object metadata, custom attributes, tags) that improve search precision and context awareness.
- Optimize semantic search at scale with sharding strategies, multi-index approaches and hierarchical indexing.
- Integrate with document management systems, knowledge bases and internal wikis.
- Keep vector stores current with incremental updates, change detection, automated synchronization and scheduled refresh pipelines.
Task 1.5: Design retrieval mechanisms for FM augmentation
- Choose document segmentation approaches: Bedrock chunking, fixed-size chunking in Lambda and custom hierarchical chunking.
- Select and configure embedding solutions such as Amazon Titan embeddings, by dimensionality and domain fit.
- Deploy vector search with OpenSearch Service, Aurora with pgvector or Bedrock Knowledge Bases.
- Improve relevance with hybrid search (keywords plus vectors) and Bedrock reranker models.
- Handle queries with query expansion, decomposition and transformation (Bedrock, Lambda, Step Functions).
- Give FMs consistent access to retrieval through function calling and Model Context Protocol (MCP) clients.
Task 1.6: Implement prompt engineering strategies and governance for FM interactions
- Control FM behavior with instruction frameworks: Bedrock Prompt Management role definitions, Guardrails and response templates.
- Build interactive systems that keep context: Step Functions clarification workflows, Comprehend intent recognition, DynamoDB conversation history.
- Govern prompts with parameterized templates, approval workflows, S3 template repositories, CloudTrail usage tracking and CloudWatch Logs.
- Test prompts for quality and regression, and refine them with structured inputs, output formats and chain-of-thought patterns.
- Design complex prompt systems with Bedrock Prompt Flows: sequential chains, conditional branching and reusable components.
Services and tools to master
- Amazon Bedrock
- Amazon Bedrock Knowledge Bases
- Amazon Bedrock Prompt Management
- Amazon Bedrock Prompt Flows
- Amazon Bedrock Guardrails
- Amazon Titan Embeddings
- Amazon OpenSearch Service
- Amazon Aurora (pgvector)
- Amazon DynamoDB
- Amazon S3
- AWS Glue Data Quality
- Amazon SageMaker AI
- AWS Lambda
- Amazon API Gateway
- AWS AppConfig
- AWS Step Functions
- Amazon Comprehend
- Amazon CloudWatch
- AWS CloudTrail
Key facts for domain 1
Short summaries; each links to the study topic with the full note and its official source.
- Chunking strategies — Fixed-size chunking sets a maximum token count and an overlap percentage; default chunking makes chunks of about 300 tokens that respect sentence boundaries.
- Hierarchical chunking — Hierarchical chunking searches small child chunks but hands the model their larger parent chunks, so it gets more context.
- Semantic chunking — Semantic chunking splits text where the meaning changes, so related sentences stay together.
- No chunking — With no chunking each document is a single chunk, so you must split the files yourself first.
- Hybrid search — Hybrid search combines vector (semantic) search with raw-text search, which helps with exact terms and acronyms that embeddings blur.
- Metadata filtering — Retrieve and RetrieveAndGenerate accept metadata filters (equals, in, greaterThan, and so on) that narrow the search before ranking.
- Reranking — A reranker model reorders the retrieved results by relevance to the query.
- Query decomposition — Query decomposition breaks a complex question into sub-queries so a multi-part question retrieves better.
- Retrieve vs RetrieveAndGenerate — Retrieve returns the source chunks; RetrieveAndGenerate also asks a model to write the answer, and RetrieveAndGenerateStream streams it.
- Citations need the output format placeholder — In a custom knowledge base prompt template, $search_results$ is required, and removing $output_format_instructions$ means the response will not contain citations.
- Titan Text Embeddings V2 — Titan Text Embeddings V2 lets you choose 1024 (default), 512 or 256 output dimensions and returns float, binary or both.
- Binary vs float vectors — Binary vectors use 1 bit per dimension versus 32 bits for float32, so they are cheaper to store but less precise.
- Reusable prompts, variants and versions — Prompt management stores reusable prompts with variables, lets you compare variants of a prompt and saves versions for deployment.
- One API for every chat model — The Converse API gives one consistent request format for all Bedrock models that support messages, so you write the code once and switch models.
- Change behavior without redeploying — AWS AppConfig changes application behavior in production without redeploying code, using feature flags or free-form configuration.
- AppConfig Agent — Applications read AppConfig data through a local endpoint exposed by the AWS AppConfig Agent, which caches the deployed configuration.
- Geographic vs global profiles — A geographic inference profile (for example US, EU or APAC) keeps processing inside that geography; a global profile routes to any supported commercial Region and costs about 10% less.
- What cross-Region inference costs and logs — There is no extra routing cost: you pay the price of the Region you call the profile from.
Guides that go deeper
Practice Domain 1
Answer practice questions, read the explanation and review your mistakes with spaced repetition.