AWS Certified AI Practitioner · AIF-C01

You can answer a thousand practice questions and still not know if you’ll pass.

CutScore reads your answers against the pass bar and tells you where you actually stand — and would rather say not yet than flatter you.

Ten questions a day, free. No card.

Your readiness · live
Projected score
678
Range 612–740 · ±64 · 96 answers · 2 hours ago
Keep going

Readiness is read off the lower bound, never the average. The band narrows only as your answers earn it.

The problem

The night before, every prep tool tells you the same thing: nothing.

You’re scoring 78% on the practice sets. So — are you ready? A practice percentage doesn’t know. The incumbents answer by selling you more questions, but more questions was never the missing piece.

The hard part of exam prep isn’t finding questions to answer. It’s knowing when you can stop.

The product

It’s a study loop, not a quiz bank.

Drill what you’re weak on, keep every miss, and watch one honest number move.

cutscore.dev/app/practice
Question 3 of 10
A team needs a foundation model to answer questions from internal product docs that change every week, without retraining. Which approach fits?
AFine-tune the model on the docs every weekWhy this is wrong: fine-tuning bakes knowledge in at training time — costly to repeat, and it still lags each update.
BRetrieval-augmented generation over the docsWhy this is correct: RAG retrieves the latest documents at query time, so edits show up immediately — no retraining.
CRaise the model’s temperature setting
DPaste every doc into the prompt each time
Correct

RAG separates the model from the knowledge: documents live in a store the model reads from at answer time, so a weekly edit is reflected the moment it lands.

Report this question

Every option carries its reasoning — why the right answer is right, and why each distractor is wrong. The source is one tap away.

Your notebook

What you’ve missed

Grouped by concept, resurfaced until you own it — and kept across every certification.

Retrieval-augmented generation

3

A team needs a model to cite internal docs that change weekly, without retraining.

Correct answer: Retrieval-augmented generation

RAG reads from a document store at query time, so updates appear without touching the model.

Mark resolved

Which component of a RAG pipeline turns documents into vectors for search?

Correct answer: An embedding model

Embeddings map text to a vector space so the retriever can find passages by meaning, not keywords.

Mark resolved

Prompt engineering

2

What does providing worked examples inside a prompt let a model do without fine-tuning?

Correct answer: Few-shot learning

A handful of in-context examples steer the output format and behaviour at inference time.

Mark resolved

Your wrong answers become a personal weakness corpus — concept-keyed, so it follows you from AIF-C01 into the next exam.

Projected score (of 1000)
678
Range 612–740 · ±64 · 96 answers · updated 2 hours ago
Keep going

The same instrument you saw up top — read off the lower bound, never the average. It says “ready” only when that bound clears 700 with margin.

01 — authored

Written, not scraped.

Every question is authored from scratch and traced to its source. No real exam items, no brain dumps.

02 — verified

Gated before you see it.

A fail-closed governance gate and a human sign-off stand between every question and you.

03 — calibrated

A number that under-claims.

Read off the lower bound, never the average. No streaks, no confetti — just the truth.

Provenance

You don’t have to take “verified” on faith.

Every published question carries its whole history. A single failed check stops it cold.

01

Blueprint task

D3.2
02

Authored from docs

official_doc
03

Sources attached

≥ 1 source
04

Governance checks

checks ✓
05

Human-approved

published
fail-closedOne missing source or one failed check, and it stays unpublished. There is no override.
Pricing

Start free. Pay when you want the whole instrument.

One all-access plan, every certification as it ships.

Free
$0

No card.

  • Ten practice questions a day
  • The daily question
  • Full concept library
  • Your readiness number, live
All-Access · best value yearly
$119/ year · or $24/mo

Everything, every cert as it ships — at one price that holds as the library grows.

  • Unlimited practice, adaptive or by domain
  • The full governed question bank
  • The Notebook, kept across certs
  • Every new certification, no upsell
Pass guaranteeIf we tell you you’re ready and you don’t pass, your next stretch is on us.
Before you ask

Plain answers.

What is the readiness number?

A projected scaled score with a confidence band that under-claims by design. We only say “ready — book your exam” when the lower bound clears the 700 pass mark with margin. On thin data, showing no number is the honest answer.

Is any of this real exam content?

Never. Every question is independently written and fully sourced. We don’t touch real exam items, brain dumps, or scraped banks.

Do I lose everything when I stop paying?

Your Notebook is yours and persists across certifications. The category’s loudest complaint is “you lose access the moment you stop paying” — we’d rather not be that.

Which exams are covered?

Today, AWS Certified AI Practitioner (AIF-C01). All-Access includes new certifications as they ship, and your concept mastery carries between them.

Stop guessing whether you’re ready.

Answer your first questions in the next two minutes.

CutScore

The exam-readiness instrument. Know if you’re ready before you book.

Company
Contact
CutScore | Know if you're ready to pass AWS AIF-C01