AWS Certified Generative AI Developer – Professional (AIP-C01) – Study Plan

Topics

• Generative AI fundamentals (LLMs, foundation models, transformer architecture)
• Key AWS AI/ML services (Bedrock, SageMaker, Rekognition, Transcribe)
• Responsible AI principles (fairness, explainability, robustness, transparency)
• Exam structure and domains

Activities

• Read the official AWS GenAI Developer exam guide.
• Create an AWS account and enable Bedrock access (request model access).
• Take ExamOS Rookie mode quiz (30 questions, 30 minutes).

Topics

• Bedrock models (Claude, Llama, Titan, Jurassic, etc.)
• Model selection criteria (cost, latency, task suitability)
• Prompt engineering for Bedrock (zero-shot, few-shot, chain-of-thought)
• API invocation via AWS SDK (boto3)

Activities

• Invoke a Bedrock model (e.g., Claude 3 Sonnet) from a Jupyter notebook.
• Experiment with different prompt templates.
• Take ExamOS Challenger mode quiz.

Topics

• Fine-tuning vs. prompt engineering vs. RAG
• Preparing training data (formatting, token limits)
• Using Amazon SageMaker JumpStart for fine-tuning
• Model evaluation (ROUGE, BERTScore, human evaluation)

Activities

• Fine-tune a small model (e.g., Flan-T5) using SageMaker.
• Compare base vs. fine-tuned outputs.
• Take ExamOS Challenger mode quiz.

Topics

• RAG architecture (retriever + generator)
• Vector databases (OpenSearch, pgvector, FAISS)
• Embedding models (Amazon Titan Embeddings, Cohere)
• Knowledge bases for Amazon Bedrock

Activities

• Build a simple RAG pipeline: ingest a PDF, chunk it, generate embeddings, query.
• Use Knowledge Bases for Amazon Bedrock.
• Take ExamOS Challenger mode quiz.

Topics

• Bedrock Agents (planning, orchestration, action groups)
• Function calling with LLMs
• Integrating Lambda functions as actions
• Multi-agent collaboration basics

Activities

• Create a Bedrock Agent that can query a weather API.
• Test the agent with natural language prompts.
• Take ExamOS Challenger mode quiz.

Topics

• Hosting custom models on SageMaker (real-time endpoints, batch transform)
• Model monitoring (drift, performance, toxicity)
• CI/CD for GenAI (SageMaker Pipelines, Model Registry)
• Cost optimization (inference pricing, provisioned throughput)

Activities

• Deploy a fine-tuned model as a SageMaker endpoint.
• Set up model monitoring for data drift.
• Take ExamOS Challenger mode quiz.

Topics

• Model security (prompt injection, adversarial attacks)
• Guardrails for Amazon Bedrock (content filters, denied topics)
• Encryption (KMS, VPC endpoints for Bedrock)
• Auditing with CloudTrail and AWS Config

Activities

• Implement Bedrock Guardrails for a sample application.
• Review AWS AI security best practices.
• Take ExamOS Challenger mode quiz.

Topics

• Multi-modal models (Claude 3 with vision, Titan Image)
• Fine-tuning with QLoRA / PEFT
• Distributed training (SageMaker Training Compiler)
• Foundation model evaluation benchmarks

Activities

• Perform parameter-efficient fine-tuning (PEFT) on a small LLM.
• Experiment with multi-modal prompts (image + text).
• Take ExamOS Challenger mode quiz.

Topics

• Full syllabus review (all exam domains)
• Time management for 75 questions (180 minutes)
• Scenario-based architecture decisions

Activities

• Take ExamOS Challenger mode full quizzes (all domains) – at least 3.
• Review every incorrect answer; study the explanation.
• Identify weak domains and retake targeted quizzes (premium Focus mode).
• Repeat until you pass 3 Challenger quizzes in a row.

Topics

• Realistic exam simulation (75 questions, 180 minutes)
• Performance-based questions (designing GenAI architectures)
• Exam day strategies

Activities

• Take ExamOS Legend mode full quizzes (80% hard questions) – at least 3.
• Review every incorrect answer – focus on "why" you missed it.
• Once you pass Legend mode twice in a row, schedule your real exam.