Has anyone passed the AWS Certified AI Practitioner exam recently?

General
1

I’m planning to take the AWS Certified AI Practitioner exam and I’m confused about which study materials are actually up-to-date and worth the time. There seem to be tons of courses, practice tests, and guides, but reviews are all over the place. Can anyone share what resources they used, how they prepared, and what the real exam focuses on so I don’t waste weeks on the wrong content

2

Took it in December. Passed. Here is what helped and what felt like a waste of time.

  1. Know what the exam focuses on
    It is entry level. Heavy on:
  • basic ML/AI concepts
  • responsible AI
  • AWS AI services and when to use which
  • a bit of GenAI and foundation models
  • pricing and high level architecture, not deep config

Almost no math. No coding.

  1. Study material that felt up to date

a) Official AWS material

  • AWS Certified AI Practitioner exam guide PDF. Start there. It defines the scope.
  • AWS sample questions for the exam on the certification page. Format matched the real exam.
  • AWS Skill Builder
    • “AWS Certified AI Practitioner Official Practice Question Set”
    • Short courses on:
    • Amazon Bedrock fundamentals
    • Amazon SageMaker foundations
    • AI services overview like Rekognition, Comprehend, Lex, Textract, Transcribe, Translate

Skill Builder content matched the wording and style of the exam better than anything else I tried.

b) Re:Invent 2023 and 2024 sessions on Bedrock and GenAI
Search for “AWS re:Invent Bedrock” on YouTube. Watch 2 or 3 recent overview sessions.
They helped with questions about:

  • what Bedrock is
  • what “foundation model” means
  • use cases and high level design
  1. Third party courses and tests

What felt useful:

  • Stephane Maarek’s AI Practitioner course on Udemy (if updated after late 2024, check last update date). Good for structure and quick review.
  • Tutorials Dojo (TD) practice questions, once they release a set specific to this exam. Their questions often feel a bit harder than the official ones, which is fine.

What felt like a waste:

  • Any course that reuses old “MLS-C01” style ML content. The pro level ML exam is different. If you see long math or hyperparameter tuning focus, skip it.
  • Random free question dumps. Many are outdated or flat out wrong. I saw questions calling Bedrock “preview” which is no longer true.
  1. What to know before you go in

Content topics I saw a lot:

  • Compare classic ML vs GenAI vs rule based
  • Explainable AI and bias
  • When to use which service:
    • Bedrock vs SageMaker vs AI services
    • Rekognition vs Textract vs Comprehend
    • Lex vs Q vs custom bot on Bedrock
  • Responsible AI, data privacy, regional data, PII
  • High level steps in an ML project lifecycle

You do not need:

  • to train models from scratch in SageMaker
  • to know exact hyperparameters
  • to write code
  1. Study path suggestion for 2 to 3 weeks

Day 1
Read the exam guide. Note each objective.

Days 2 to 5
AWS Skill Builder AI Practitioner official path
Take notes on:

  • definitions
  • which service for which use case
  • basic pricing ideas

Days 6 to 8
Watch 2 or 3 Bedrock and GenAI overview talks on YouTube.
Review AI responsibility / bias content from AWS.

Days 9 to 11
Do official AWS practice questions.
Do 1 or 2 third party practice sets.
Review every wrong answer. Learn why each wrong option is wrong, not only why the right one is right.

Days 12 to 14
Light review of notes. Rewatch any weak topics, for me that was responsible AI and some service edge cases.

  1. Exam day tips
  • Expect many “which service fits this scenario” questions.
  • Eliminate obviously wrong services first.
  • Watch the wording. They love phrases like “minimal ML expertise” or “need for customization” to steer you toward AI services or SageMaker.

If you already know general AI concepts and a bit of AWS, 1 to 2 weeks focused on the official AWS content plus one practice set is enough. If you are new to both AI and AWS, plan closer to 3 to 4 weeks of short daily study sessions.

3

Passed it in early Jan. Co-signing a lot of what @sterrenkijker said, but I’d tweak the strategy a bit so you don’t over-study the wrong stuff.

1. What’s actually current & worth it

If you want to minimize fluff:

  • Official exam guide & sample questions: Non‑negotiable. Use the blueprint as a checklist and ignore anything not clearly mapped to it.
  • Skill Builder: Good, but I’d prioritize:
    • Bedrock fundamentals
    • “AI services overview” style content
    • Responsible AI & governance
      Skip deep SageMaker labs if you’re short on time. They’re cool, but overkill for this exam.

Where I slightly disagree with @sterrenkijker: you don’t need a ton of re:Invent videos. One or two recent Bedrock overviews are enough. People get lost binging 10+ talks and then get hit with basic “which service should you choose” questions.

2. Extra resources that didn’t waste my time

  • AWS docs “What is X?” pages for:

    • Bedrock
    • SageMaker (high level sections only)
    • Rekognition, Comprehend, Textract, Transcribe, Translate, Lex, Kendra, Amazon Q
      Just read the top sections: what it is, main use cases, and key features. No need to deep dive every API.

  • One third‑party course + one practice set

    • I used a Udemy course as a structured review, not a primary source. If a section didn’t match the exam guide language, I mentally downgraded it.
    • Practice questions: do fewer, but review them harder. One decent set > five garbage dumps.

3. Topics people surprisingly ignore

These came up more than I expected:

  • Data privacy & region considerations
    Things like “keep data in-region,” “don’t store PII unnecessarily,” encryption-at-rest vs in-transit, which service is managed vs custom.

  • Limits of GenAI vs classic ML
    When GenAI is overkill, when a simple rules engine or standard ML model is enough, hallucination risk, evaluation / human oversight.

  • Cost & effort tradeoffs
    If the question says:

    • “Minimal ML experience” or “fastest time to value” → managed AI services
    • “Custom model, specific domain, more control” → SageMaker
    • “Use multiple FMs with minimal infra work” → Bedrock

4. Stuff I wouldn’t spend time on

  • Long math, derivations, tuning algorithms
  • Full-blown SageMaker pipelines and MLOps
  • Trying to memorize every individual Bedrock model provider and their marketing fluff
  • Sketchy “brain dumps” that look like they were written three services ago (if they say Bedrock is preview, close the tab)

5. Quick sanity check for your prep

You’re probably ready if you can, without notes:

  • Explain in plain english: GenAI vs traditional ML vs rules-based.
  • Match scenarios to:
    • Bedrock vs SageMaker vs a managed AI service.
  • Answer:
    • “You need document text & structure” → Textract
    • “You need sentiment / entities in text” → Comprehend
    • “You need speech to text” → Transcribe
    • “You need image / video analysis” → Rekognition
    • “You need a chatbot with intents” → Lex
    • “You need search/Q&A over docs” → Kendra / Amazon Q depending on context.

If you can also talk about bias, explainable AI at a high level, and data governance basics, you’re in the zone.

tl;dr: Lean on the official stuff, skim just enough docs to know when to use which service, one solid course + one practice set, and avoid anything that smells like the old pro-level ML exam.