AWS AI Practitioner study plan: 2-week and 4-week schedules

Here's a concrete, day-by-day plan for the AWS Certified AI Practitioner (AIF-C01): a 4-week track if AI and AWS are both new to you, and a 2-week track if you already know basic AWS or work around machine learning. Follow either one and finish consistently scoring 800+ on practice exams before test day.

The AI Practitioner is a concepts exam, not a coding exam. You won't train a model or write Python. You will need to recognize what each AWS AI service does, and to make good judgment calls: when to use RAG instead of fine-tuning, which inference setting makes output more creative, and which service detects bias. This plan is built around those calls.

The exam you're planning for

The two generative-AI domains, Fundamentals of Generative AI and Applications of Foundation Models, carry 52% of the exam between them. Both plans below spend the most time there. Besides standard multiple choice, expect multiple-response, ordering and matching questions.

Pick your track

The daily hour (both tracks)

Every study day follows the same shape:

• 20 minutes — learn: one topic from that day's list (AWS docs, a course video, or your notes).
• 30 minutes — practice questions on that topic, plus a few review questions from earlier days.
• 10 minutes — review misses: read the explanation for every wrong answer and say out loud why the right answer wins and the others don't.

The 4-week plan (beginner track)

Week 1 — Fundamentals of AI and ML (20%)

• Days 1–2: AI vs ML vs deep learning; supervised, unsupervised and reinforcement learning; classification vs regression vs clustering.
• Day 3: The ML lifecycle — training/validation/test splits, features and labels, overfitting vs underfitting, batch vs real-time inference. Metrics: accuracy, precision, recall, F1 for classification; RMSE for regression.
• Day 4: The managed AI services by job: Rekognition (images/video), Textract (documents), Comprehend (text/sentiment), Transcribe and Polly (speech in/out), Translate, Lex (chatbots), Personalize (recommendations), Kendra (enterprise search).
• Day 5: Amazon SageMaker at a "what is it for" level, plus Ground Truth for labeling.
• Weekend: A 25-question quiz over the whole domain. Anything under ~70%, re-read that topic.

Week 2 — Fundamentals of Generative AI (24%)

• Days 1–2: Foundation models, tokens, embeddings, transformers, the context window, multimodal and diffusion models. What hallucination is and why it happens.
• Day 3: Inference parameters — temperature, top-p, max tokens, stop sequences — and what each does to the output.
• Day 4: Amazon Bedrock: one API to models from several providers, on-demand (per-token) vs Provisioned Throughput pricing. Also PartyRock, SageMaker JumpStart, Amazon Q Business and Amazon Q Developer.
• Day 5: Generative AI's strengths and limits (adaptability vs accuracy, interpretability, cost) — the exam likes "which is a disadvantage of…" questions.
• Weekend: Domain quiz plus a re-drill of misses.

Week 3 — Applications of Foundation Models (28%)

• Day 1: Prompt engineering — zero-shot, few-shot, chain-of-thought, prompt templates, negative prompting.
• Day 2: RAG end to end: chunking, embeddings, vector stores (OpenSearch, Aurora PostgreSQL with pgvector), and Knowledge Bases for Amazon Bedrock.
• Day 3: Customizing a model: fine-tuning (labeled prompt/completion pairs) vs continued pre-training (unlabeled domain text) vs just prompting or RAG. Use the cheat sheet below until the choice is automatic.
• Day 4: Agents for Amazon Bedrock and action groups; prompt injection and jailbreaking; Guardrails for Amazon Bedrock.
• Day 5: Evaluating models: ROUGE (summaries), BLEU (translation), human evaluation, Bedrock Model Evaluation, and business metrics vs model metrics.
• Weekend: Mixed quiz across the first three domains.

Week 4 — Responsible AI, security, and practice exams

• Day 1: Responsible AI (14%): fairness and bias, explainability, transparency, toxicity, model cards and AI Service Cards. Services: SageMaker Clarify (bias, explainability), Model Monitor (drift), Augmented AI (human review).
• Day 2: Security, compliance and governance (14%): IAM roles and least privilege, KMS, Macie, PrivateLink, CloudTrail, Config, Artifact, Audit Manager, the shared responsibility model and data residency.
• Day 3: Full timed 65-question mock exam. Score it, list every miss by domain.
• Day 4: Drill your two weakest domains, then a second full mock. You want 800+.
• Day 5: A third mock only if day 4 was under 800; otherwise light review.
• Exam day −1: Light review only — the cheat sheet below, the AI services by job, and the responsible-AI services. Stop early. Sleep.

Cheat sheet: prompt, RAG, or fine-tune?

This one decision shows up again and again on the AIF-C01, phrased a dozen different ways:

When two answers both seem to work, the exam usually wants the lowest-cost, lowest-effort option that meets the requirement. That's why RAG beats daily fine-tuning for "always up to date" scenarios.

The 2-week plan (experienced track)

Same structure, compressed. Take each domain's quiz first and only study what you miss.

• Days 1–2: AI/ML fundamentals and the managed AI services by job.
• Days 3–5: Generative AI fundamentals, Bedrock, inference parameters, Amazon Q. Heavy question volume — 40+ per day.
• Days 6–8: Applications of foundation models: prompting, RAG, fine-tuning vs pre-training, agents, guardrails, evaluation metrics. This is the biggest domain — don't rush it.
• Day 9: Responsible AI and security/governance. Experienced people most often lose points here, because it's AWS-specific service names you may never have needed at work.
• Day 10: Full timed mock #1. List misses by domain.
• Days 11–12: Drill weak domains.
• Day 13: Full timed mock #2. Book the real exam once you clear 800.
• Day 14: Light review, early night.

The final 48 hours (both tracks)

• Two days out: last full mock. If it's 800+, you're done learning — trust it.
• One day out: a 20-minute skim of the cheat sheet above, the AI services by job, and Clarify / Model Monitor / Augmented AI. No new topics, no full exams.
• Exam day: flag anything that takes over 90 seconds and come back. 90 minutes for 65 questions is generous, and second-pass questions often click instantly.

After you pass

If you skipped it, the Cloud Practitioner is a quick add — you'll already know a good share of its security and global-infrastructure material. If you want a role-based AWS certification next, the Solutions Architect Associate is the one employers shortlist for. Not sure which direction fits? Take the two-minute which certification should I take quiz.

Frequently asked questions