AWS AI Practitioner (AIF-C01) Master Resource Guide
The ultimate study guide for the AWS AI Practitioner (AIF-C01) exam. Learn the custom matrix, Bedrock vs SageMaker, and secure your certification.
As a student looking for a clear path to acing AI practitioner certification, here is the master blueprint and reference guide for the community of everything I've compiled. It strips away the noise, links directly to the official resources within the AWS Skill Builder ecosystem, and outlines the systems-level decision matrices required to pass.
Here is a master list of resources to help those who are interested in the new AWS Certified AI Practitioner exam.
⭐ If you find this guide useful, please bookmark and upvote it so other learners can locate it easily.
- To help each other out, we can comment below with our findings and progress on the certification, questions, practice, as well as tips to achieve our objectives together!
- Also, if you have any suggestions to add, I'm open to listening to your ideas!
Resource links:
- Exam Details:
The exam code is AIF-C01
Official AWS Certification page on AIF
The first resource to usually read is the Exam Guide as that tells you what is in / out of scope.
- Learning plans:
- Practice Exams:
- Gamified learning:
Cloud Quest - Generative AI is a game based learning platform - provides a digital badge on completion.
Free "Card Clash" Generative AI Game teaches you how the various Generative AI related services work together.
Digital classrooms:
Exam booster:
Other optional resources:
1. Exam Blueprint & Critical Boundaries
| Attribute | Exam Specification |
|---|---|
| Exam Code | AIF-C01 (Foundational Level) |
| Questions / Time | 65 questions (multiple choice, multi-select, ordering) in 90 minutes |
| Scoring | Scaled score: 100–1000. Passing score: 700 (15 unscored diagnostic questions) |
| Official Exam Hub | AWS Certified AI Practitioner Official Portal |
| Official Exam Guide | AIF-C01 Detailed Exam Guide PDF |
The Five Scored Domains
- Domain 1: Fundamentals of AI and ML (20%) — ML types (supervised/unsupervised/reinforcement), training vs. inference, evaluation metrics (precision, recall, F1, RMSE).
- Domain 2: Fundamentals of Generative AI (24%) — LLMs, embeddings, tokens, temperature, context windows, prompt patterns, hallucination mitigation.
- Domain 3: Applications of Foundation Models (28%) — Bedrock vs. SageMaker, RAG vs. Fine-tuning, Bedrock Agents, multi-modal workflows.
- Domain 4: Guidelines for Responsible AI (14%) — Bias mitigation, fairness metrics, explainability, safety guardrails, human-in-the-loop workflows.
- Domain 5: Security, Compliance, and Governance (14%) — IAM access control, KMS customer-managed keys, CloudTrail auditability, data privacy compliance.
Out of Scope Boundary: You do not need to write Python algorithms from scratch, calculate linear algebra/backpropagation by hand, or manage distributed Kubernetes/EKS clusters. The exam tests architectural trade-offs and service selection.
2. The Official AWS Skill Builder Learning Path
You do not need to buy external slide decks. Follow this direct progression using AWS Skill Builder:
Phase 1: Core Curriculum & Hands-On Sandboxes
- Curated Exam Category: Access the full collection via the Skill Builder AIF-C01 Exam Prep Hub .
- Digital Learning Plan: Enroll in the AWS AI Practitioner Learning Plan for direct coverage across all five domains.
- Interactive RPG Practice: If you prefer experiential learning over slides, play AWS Cloud Quest: Generative AI . You solve problems inside live AWS console sandboxes (deploying Bedrock models, configuring Knowledge Bases, and setting guardrails) without accumulating personal billing costs. Finishing it earns an official Credly digital badge.
Phase 2: Practice Assessments & Calibration
- Official 20-Question Diagnostic Set: Launch the Official Practice Question Set: AWS Certified AI Practitioner on AWS Skill Builder to calibrate question cadence and wording.
- Full-Length Timed Simulation: Complete the Official Practice Exam (AIF-C01) on Skill Builder to simulate the actual 90-minute testing environment.
- Third-Party Review Mode (Optional): If you want a secondary independent set with detailed per-question distractor explanations, use the Tutorials Dojo AIF-C01 Practice Exams .
- Strict Warning: Avoid unauthorized brain dump sites. Using brain dumps violates the AWS Certification Candidate Agreement and results in certification revocation and multi-year testing bans.
3. The Architecture Decision Matrix (Cheat Sheet)
AIF-C01 scenario questions place multiple valid-looking services in the choices. Use these decision trees to identify the right answer immediately.
A. The Foundation Model Customization Ladder
1
2
3
4
5
6
7
8
9
10
11
12
13
[ How should the team adapt the Foundation Model? ]
│
├─► Zero training cost? Fast formatting or persona change?
│ └─► PROMPT ENGINEERING (System prompts, few-shot examples)
│
├─► Needs private internal data, real-time updates, or exact source citations?
│ └─► RAG (Retrieval-Augmented Generation via Amazon Bedrock Knowledge Bases)
│
├─► Model must speak custom company jargon, industry acronyms, or output rigid JSON?
│ └─► FINE-TUNING (Labeled input-output dataset; updates model weights)
│
└─► Unique non-English dialect or building a model from scratch?
└─► PRE-TRAINING / CONTINUED PRE-TRAINING (SageMaker HyperPod; high budget)
B. Managed Bedrock vs. Custom SageMaker
| Architectural Need | Correct AWS Service | Key Qualifying Phrase in Exam Question |
|---|---|---|
| Fully Serverless GenAI API | Amazon Bedrock | "Zero infrastructure management", "API-driven Foundation Models", "Claude/Llama/Nova access" |
| Dynamic Vector Search / RAG | Bedrock Knowledge Bases | "Connect private documents in S3 to a foundation model without fine-tuning" |
| Autonomous Workflow Invocation | Amazon Bedrock Agents | "Multi-step task completion using dynamic Lambda functions and API tools" |
| Real-time LLM Guardrails & PII Masking | Amazon Bedrock Guardrails | "Prevent prompt injection, block toxic keywords, and redact social security numbers in real-time chats" |
| Custom Model Training & Containers | Amazon SageMaker AI | "Deploy custom Docker container", "hyperparameter tuning jobs", "manage dedicated GPU compute" |
| Demographic Bias & Explainability | Amazon SageMaker Clarify | "Inspect training data for disparate impact", "compute SHAP values for model transparency" |
| Model Governance & Auditing | SageMaker Model Cards | "Document model intended use, limitations, and performance metrics for regulatory compliance" |
| No-Code Tabular Machine Learning | SageMaker Canvas | "Business analyst predicting customer churn without writing code" |
| Human Validation Loop | Amazon Augmented AI (A2I) | "Route predictions below 80% confidence to human annotators for review" |
4. Distinguishing Lookalike AWS Services
- Amazon Q Business vs. Amazon Bedrock:
- If an enterprise wants a ready-to-deploy intranet assistant connected to Microsoft 365, Salesforce, and Google Drive without writing code, pick Amazon Q Business.
- If developers are writing code to build a proprietary generative AI application, pick Amazon Bedrock.
- Amazon Q Developer vs. Amazon CodeGuru:
- Amazon Q Developer: Interactive AI coding companion (generates functions, writes unit tests, explains code, updates Java runtimes).
- Amazon CodeGuru: Static automated code analysis for detecting security bugs, resource leaks, and CPU profiling bottlenecks.
- Amazon Comprehend vs. Amazon Bedrock:
- Amazon Comprehend: Specialized, pre-trained traditional NLP service (sentiment classification, entity extraction, syntax analysis, PII identification).
- Amazon Bedrock: Foundation model platform that handles complex generative reasoning and open-ended text synthesis.
- Amazon Transcribe / Polly vs. Textract / Rekognition:
- Transcribe: Speech to text.
- Polly: Text to speech.
- Textract: OCR document extraction (retaining table structures and form keys).
- Rekognition: Computer vision (detecting objects, faces, celebrities, and inappropriate visual scenes).
5. Exam Day Execution Strategy
- Answer Every Question: There is no negative marking or score deduction for wrong answers. Never leave a question blank.
- Watch the Constraints: When reading questions, immediately underline constraint terms: "at the lowest operational cost", "without managing servers", or "with minimal application development". These phrases eliminate complex answers in favor of managed solutions like Bedrock or Q.
- Pacing: Budget 1 minute per question on your first pass. Flag questions where you are split between two options and return to them during the final 25 minutes.
⭐ If you find this guide useful, please bookmark and upvote it so other learners can locate it easily.
- To help each other out, we can comment below with our findings and progress on the certification, questions, practice, as well as tips to achieve our objectives together!
- Also, if you have any suggestions to add, I'm open to listening to your ideas!
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article