AWS AI Certification Study Guide 2026: AIF-C01, MLA-C02, AIP-C01

By Kyle Tice · Updated October 19, 2026 · Facts checked against official exam pages

If you’re new to AI on AWS, start with the AI Practitioner (AIF-C01). If you build on AWS, go to the Machine Learning Engineer – Associate (MLA-C02) next, then finish with the Generative AI Developer – Professional (AIP-C01). If you’re a manager or consultant who will never touch the console, take the new AI Business Strategist (AIB-C01) and stop there.

This guide covers the order, the official domain weights, a week-by-week plan for each exam, the free and paid resources worth your time, and the labs that actually prepare you.

What changed in 2026

The AWS AI track changed a lot this year. Check these before you book anything.

• The ML Specialty (MLS-C01) is retired. The last day to sit it was March 31, 2026. If you hold it, it stays valid until it expires. AWS now points ML candidates to MLA and AIP. See /certifications/aws-ml-specialty/.
• MLA-C01 is being replaced by MLA-C02. The English C01 exam was available through September 28, 2026. The C02 beta started September 29, 2026 at $75 (beta pricing), English only, 85 questions in 170 minutes. AWS says no domains were added, but C02 adds Amazon Bedrock, RAG, agentic AI, foundation models and responsible AI to the scope. The general availability date is still TBD on AWS’s page.
• AIP-C01 is live. The Generative AI Developer – Professional is a 75-question, 180-minute, $300 exam with a 750 passing score.
• AIB-C01 is new and in beta. The AI Business Strategist is a non-technical exam. Beta price is $50 (standard $100). If you earn it by February 15, 2027, you also get an Early Adopter badge.
• AIF-C01 language changes. AWS retired the Italian and German versions of AIF-C01 after October 15, 2026.

Which AWS AI exam to take, and in what order

Prices as of October 2026: AIF-C01 $100, AIB-C01 $100 ($50 beta), MLA-C02 $150 ($75 beta), AIP-C01 $300. All AWS certifications in this family are valid for three years.

Who should skip: If you already build RAG apps on Bedrock at work, AIF-C01 adds little. Book AIP-C01. If you don’t work in an AWS shop, a vendor-neutral portfolio project will do more for you than any of these; see /guides/ai-portfolio-projects/.

AIF-C01: AI Practitioner

Exam domains (official)

65 questions (50 scored, 15 unscored), 90 minutes, passing score 700/1000. Question types include multiple choice, multiple response, ordering and matching. Coding is explicitly out of scope.

3-week plan (about 20–40 hours)

• Week 1: Domains 1 and 2. Supervised vs. unsupervised learning, model evaluation metrics, transformers, tokens, embeddings, context windows. Watch the free AWS Skill Builder exam prep plan.
• Week 2: Domain 3, the largest. Prompt engineering techniques, RAG vs. fine-tuning vs. continued pre-training, Bedrock Knowledge Bases, Bedrock Agents, model selection factors (cost, latency, context length). Do the labs below.
• Week 3: Domains 4 and 5. Bedrock Guardrails, SageMaker Clarify, model cards, IAM, the shared responsibility model, data privacy. Take the free official practice set, then one paid set. Book the exam when you score above 80% on unseen questions.

MLA-C02: Machine Learning Engineer – Associate

Exam domains (official, C02)

The C02 exam guide lists 50 scored and 15 unscored questions with a passing score of 720. The beta is longer (85 questions, 170 minutes) because it carries more unscored items. AWS recommends at least a year with SageMaker AI and Bedrock, plus a year in a related role.

What C02 adds: Bedrock features, FM fine-tuning, RAG architectures, agentic workflows and their observability, and responsible AI. If your study material says “MLA-C01” and never mentions Bedrock, it’s out of date.

8-week plan (about 60–100 hours)

• Weeks 1–2: Data prep (28%). S3, Glue, Athena, Data Wrangler, SageMaker Feature Store, data formats (Parquet, RecordIO), handling imbalance and missing values, PII handling.
• Weeks 3–4: Model development. SageMaker built-in algorithms (XGBoost, Linear Learner), training jobs, Automatic Model Tuning, evaluation metrics, SageMaker Model Registry. Then the FM half: choosing Bedrock models, fine-tuning vs. RAG, evaluation.
• Weeks 5–6: Deployment and orchestration. Real-time, serverless, async and batch inference, autoscaling, SageMaker Pipelines, CodePipeline, infrastructure as code, agentic workflow orchestration.
• Week 7: Monitoring and security. Model Monitor, drift, CloudWatch, CloudTrail, VPC endpoints, KMS, IAM least privilege.
• Week 8: Practice exams and gap review.

AIP-C01: Generative AI Developer – Professional

Exam domains (official)

75 questions (65 scored, 10 unscored), 180 minutes, passing score 750. AWS recommends two or more years building production applications and one year of hands-on GenAI work. Model training and feature engineering are out of scope; this is an application developer’s exam.

8–10 week plan (about 80–120 hours)

• Weeks 1–3: Domain 1. Vector stores (OpenSearch, Aurora pgvector), chunking and embedding strategy, Bedrock Knowledge Bases, prompt management, data compliance.
• Weeks 4–5: Domain 2. Agents (Bedrock Agents and AgentCore), tool use, API integration, streaming, event-driven designs with Lambda and Step Functions.
• Week 6: Domain 3. Guardrails, prompt-injection defenses, PII redaction, IAM for agents, audit logging.
• Week 7: Domains 4 and 5. Token cost control, caching, provisioned throughput vs. on-demand, latency tuning, model evaluation, troubleshooting bad retrieval.
• Weeks 8–10: Practice exams and rebuilding weak areas as small projects.

AIB-C01: AI Business Strategist

Exam domains (official)

The exam guide lists a 130-minute duration and a 700 passing score; AWS’s certification page lists the beta at 85 questions and 170 minutes. No coding or hands-on AWS work is expected. You need business-level familiarity with Bedrock, SageMaker AI, Amazon Quick, the AWS Cloud Adoption Framework, the Well-Architected Responsible AI Lens and AWS pricing tools.

3-week plan (about 20–40 hours)

• Week 1: AI literacy and solution types. If you’ve passed AIF-C01, skim this.
• Week 2: Strategy and value: KPIs, ROI frameworks, build-buy-partner decisions, pricing models (consumption, instance, seat).
• Week 3: Governance, risk, change management, scaling from pilot to production. Read the Responsible AI Lens.

Free official resources first

1. AWS Skill Builder (skillbuilder.aws). Each exam has an exam prep plan, and official practice question sets are available. Start here for every exam.
2. The exam guides themselves (linked in Sources). Read the in-scope and out-of-scope services lists. They tell you what you can ignore.
3. AWS documentation and workshops. The Bedrock user guide and the SageMaker AI developer guide answer most “which service” questions.
4. Free video courses. Andrew Brown’s ExamPro courses on freeCodeCamp’s YouTube channel are a good free option for AIF-C01. Confirm the course matches the current exam version before you commit hours.

Paid resources worth the money

• Video courses. Stephane Maarek’s AWS courses on Udemy are the most common paid choice for AIF-C01, and his MLA course (with Frank Kane) covers the associate exam. Check that the course has been updated for C02 before buying. Browse Stephane Maarek’s courses on Udemy.
• Practice exams. Tutorials Dojo sells practice sets for AIF, MLA and AIP, with detailed explanations. Their AIP-C01 set includes a free 30-question sampler. See Tutorials Dojo practice exams.
• Alternate question bank. Whizlabs offers AWS practice tests and some hands-on labs if you want a second question pool. Browse Whizlabs AWS practice tests.

You don’t need all three. For AIF-C01, one course plus one practice set is plenty. For AIP-C01, spend money on practice exams, not a second video course, and spend the rest of your time building.

Practice-exam strategy

• Take one practice test cold at the start. Use the per-domain results to decide where your hours go.
• Read every explanation, including for questions you got right. AWS questions turn on one keyword (“least operational overhead”, “most cost-effective”). Learn to spot it.
• Track scores by domain, not overall. A 78% overall can hide a 50% in Domain 1 of AIP, which is 31% of the exam.
• Don’t memorize question banks. Practice sets get recycled. If you’re scoring 95% on a set you’ve seen twice, that number means nothing. Switch to a fresh set.
• Never use exam dumps. They violate the AWS certification agreement and can get your credential revoked.
• Beta exams are different. On MLA-C02 and AIB-C01 betas, expect longer exams with unscored items, results that take longer than usual, and little third-party practice material that matches the new content. Lean harder on the official exam guide.

Hands-on labs to do

Use a personal AWS account with a billing alarm set first. Most of these cost a few dollars if you tear them down the same day.

For AIF-C01

1. Enable model access in Amazon Bedrock and compare two models in the playground on the same prompt. Change temperature and top-p and note what changes.
2. Build a Bedrock Knowledge Base over a few PDFs and ask it questions.
3. Add a Bedrock Guardrail that blocks a denied topic and redacts PII.

For MLA-C02

1. Train and tune an XGBoost model in SageMaker AI on a public tabular dataset, register it, and deploy it to a serverless endpoint.
2. Build a SageMaker Pipeline with processing, training, evaluation and a conditional register step.
3. Turn on Model Monitor and create drift by sending skewed data.
4. Fine-tune or customize a Bedrock model on a small dataset and compare it to a RAG approach.

For AIP-C01

1. Build a RAG app end to end: chunk, embed, store in OpenSearch Serverless or Aurora pgvector, retrieve, generate, cite sources.
2. Build an agent with two tools (one Lambda function, one API) and add guardrails.
3. Write a small evaluation set (20–50 questions with expected answers) and run it after every prompt change.
4. Measure token cost and latency for one workload, then cut both with caching or a smaller model.

These projects also make good portfolio pieces. Hiring managers look at your GitHub, not just your badge.

Related reading

• AIF-C01 vs AI-901 and AIF-C01 vs Google Generative AI Leader
• AWS MLA vs Google PMLE
• AI-103 vs AIP-C01 and Claude Architect vs AWS GenAI Developer
• Current vouchers and discounts and the cost calculator

FAQ

Should I take MLA-C02 now while it’s in beta?

If you’re already prepared, yes: the beta costs $75 instead of $150 and passes count as full certifications. The tradeoffs are a longer exam, slower results and fewer up-to-date practice materials. If you’re just starting, study against the C02 guide anyway, since that’s the version you’ll sit.

Is the ML Specialty still worth anything?

If you hold it, it stays valid until it expires and still reads well on a résumé. You can’t earn it anymore. For a current credential, MLA-C02 is the closest replacement and AIP-C01 covers the GenAI side.

Do I need AIF-C01 before AIP-C01?

No. AWS has no prerequisites for these exams. AIF-C01 is a reasonable warm-up if you’re new to AWS AI services, but experienced developers can go straight to AIP-C01.

How long are AWS AI certifications valid?

Three years. To renew, you retake the current version of the exam (or, for AIF-C01, earning MLA also renews it, per AWS’s renewal rules).

Which one do employers ask for most?

The AI Practitioner is the most common entry-level AWS AI credential, and the ML Engineer – Associate is the most relevant for hands-on ML roles. For production GenAI roles, AIP-C01 plus a public RAG or agent project is a strong combination. See /rankings/ for how we score demand.