Cheat Sheet & Exam Strategy

Every high-yield AIF-C01 association on one page, plus how to attack scenario questions on exam day.

10 min read

Concept associations

When you see…Think…
Labeled dataSupervised learning
No labels, find groupsUnsupervised learning / clustering
Reward, trial and errorReinforcement learning
Predict a category / a numberClassification / regression
Great on training data, bad on new dataOverfitting
Can't miss a positive caseOptimize recall
False alarms are costlyOptimize precision
Imbalanced datasetDon't trust accuracy; use F1/AUC
Confidently wrong answersHallucination → fix with RAG + human review
Same prompt, different outputsNondeterminism → lower temperature
Creativity dialTemperature
Examples inside the promptFew-shot (in-context learning)
"Think step by step"Chain-of-thought
Malicious instructions in inputPrompt injection
Company documents + citations, data changes oftenRAG / Knowledge Bases
Change model behavior/style with labeled pairsFine-tuning
Unlabeled domain corpusContinued pre-training
Multi-step tasks calling APIsBedrock Agents
Filter harmful content / mask PII in GenAIBedrock Guardrails
Summarization metric / translation metricROUGE / BLEU
Bias detection & explainabilitySageMaker Clarify
Drift in productionSageMaker Model Monitor
Human review of predictionsAmazon A2I
Document your modelSageMaker Model Cards
Keep GenAI traffic off the internetPrivateLink / VPC endpoints
Who invoked the model (audit)CloudTrail
Configuration complianceAWS Config

Service one-liners

ServiceOne line
Amazon BedrockFMs from many providers via one serverless API
Bedrock Knowledge BasesManaged RAG over your documents
Bedrock AgentsLLM that plans steps and calls your APIs
Bedrock GuardrailsSafety filters + PII masking for GenAI
Amazon Q BusinessReady-made employee assistant on company data
Amazon Q DeveloperAI coding assistant
AWS PartyRockFree no-code GenAI playground
Amazon SageMaker AIBuild/train/deploy custom ML
SageMaker JumpStartPre-built models to start from
SageMaker ClarifyBias detection + explainability
SageMaker Model MonitorProduction drift monitoring
SageMaker Ground TruthHuman data labeling
Amazon RekognitionImages & video analysis
Amazon ComprehendText NLP: sentiment, entities, PII
Amazon Transcribe / PollySpeech→text / text→speech
Amazon TranslateTranslation
Amazon TextractExtract text/forms from documents
Amazon LexChatbots
Amazon KendraIntelligent enterprise search
Amazon Personalize / ForecastRecommendations / time-series forecasts
Amazon OpenSearchVector database for semantic search
Trainium / InferentiaAWS chips for training / inference

Exam-day tactics

  1. Find the constraint: "no ML expertise" → AI service; "lowest cost, no training" → prompting/RAG; "must cite sources" → RAG; "consistent behavior" → fine-tuning.
  2. Eliminate the extremes: "train from scratch" and answers removing human oversight are nearly always wrong.
  3. Watch superlatives: LEAST effort → most managed option; MOST cost-effective → smallest sufficient model / prompting before tuning.
  4. Budget ~80 seconds per question; flag and return rather than stall.
  5. Answer everything — no penalty for guessing.
Tip

Take Mock Exam 1 now. Review misses, reread those lessons, drill the practice bank, then confirm with Mock Exam 2. Two 750+ scores = book the real exam with confidence.

Knowledge check
Question 1 of 3

Match the need: "Our chatbot must answer from current company documents AND show its sources, with no model training."