Domain 1: Fundamentals of AI and ML

AWS Certified AI Practitioner — Domain 1: Fundamentals of AI and Machine Learning

Exam Code: AIF-C01  |  Level: Foundational
Domain Weight: 20%  |  Total Domains: 5  |  Passing Score: 700/1000

---

Table of Contents

1. AI, ML, and Deep Learning
   • 1.1 Definitions and Hierarchy
   • 1.2 Types of AI by Capability
2. Types of Machine Learning
   • 2.1 Supervised Learning
   • 2.2 Unsupervised Learning
   • 2.3 Reinforcement Learning
   • 2.4 Semi-Supervised and Self-Supervised Learning
3. The ML Lifecycle
   • 3.1 Pipeline Stages
4. Key ML Concepts and Terminology
   • 4.1 Bias, Variance, and Regularization
   • 4.2 Hyperparameters vs. Parameters
5. Data Fundamentals for ML
   • 5.1 Data Types and Quality
   • 5.2 Handling Missing Data and Class Imbalance
6. Feature Engineering
   • 6.1 Feature Types and Transformations
7. Model Training and Evaluation
   • 7.1 Evaluation Metrics
8. ML Algorithms Overview
   • 8.1 Supervised Algorithms
   • 8.2 Unsupervised Algorithms
9. Deep Learning and Neural Networks
   • 9.1 Network Architectures
10. AWS AI and ML Services
    • 10.1 Amazon SageMaker
    • 10.2 AWS Pre-Built AI Services
11. Exam Tips and Quick Reference

---

1. AI, ML, and Deep Learning

Artificial Intelligence, Machine Learning, and Deep Learning form a nested hierarchy where each is a subset of the one above it. Understanding where each term begins and ends is the foundation for the entire exam.

1.1 Definitions and Hierarchy

Artificial Intelligence (AI) is the simulation of human intelligence in machines programmed to think, reason, and make decisions. Machine Learning (ML) is a subset of AI where models learn patterns from data without being explicitly programmed with rules. Deep Learning (DL) is a subset of ML using multi-layered neural networks.

┌──────────────────────────────────────────┐
│          Artificial Intelligence         │
│  ┌────────────────────────────────────┐  │
│  │        Machine Learning            │  │
│  │  ┌──────────────────────────────┐  │  │
│  │  │       Deep Learning          │  │  │
│  │  │  ┌────────────────────────┐  │  │  │
│  │  │  │    Generative AI       │  │  │  │
│  │  │  └────────────────────────┘  │  │  │
│  │  └──────────────────────────────┘  │  │
│  └────────────────────────────────────┘  │
└──────────────────────────────────────────┘

Key Concept: All ML is AI, but not all AI is ML. All Deep Learning is ML, but not all ML is Deep Learning. Generative AI is a subset of Deep Learning.

1.2 Types of AI by Capability

Exam Tip: Every AWS AI service — Bedrock, SageMaker, Rekognition — is narrow AI. AGI and ASI are conceptual terms, not products.

---

2. Types of Machine Learning

ML is categorized by how models learn from data. The exam frequently gives a scenario and asks you to identify the correct learning paradigm.

2.1 Supervised Learning

In supervised learning, the model learns from labeled data — datasets where the correct answer is already known. The goal is to learn a mapping from inputs to outputs.

Classification

Predicts a discrete class or category. Two sub-types exist:

Regression

Predicts a continuous numeric value. Examples include house price prediction and temperature forecasting.

2.2 Unsupervised Learning

The model finds patterns in unlabeled data. There is no correct answer to learn from — the model discovers structure on its own.

2.3 Reinforcement Learning

An agent learns by interacting with an environment, receiving rewards for good actions and penalties for bad ones. It learns a policy — a strategy for maximizing cumulative reward over time.

Exam Tip: If a scenario describes a system learning through trial, reward, and penalty — game playing, robotic control, autonomous driving — the answer is reinforcement learning.

2.4 Semi-Supervised and Self-Supervised Learning

---

3. The ML Lifecycle

The ML lifecycle is an iterative, six-stage process from business problem definition to production monitoring. Projects routinely loop back through earlier stages when performance degrades or requirements change.

3.1 Pipeline Stages

┌──────────────┐    ┌──────────────┐    ┌──────────────┐
│  1. Define   │───►│  2. Collect  │───►│  3. Prepare  │
│   Problem    │    │    Data      │    │    Data      │
└──────────────┘    └──────────────┘    └──────────────┘
                                               │
┌──────────────┐    ┌──────────────┐    ┌──────▼───────┐
│  6. Monitor  │◄───│  5. Deploy   │◄───│  4. Train &  │
│  & Retrain   │    │    Model     │    │   Evaluate   │
└──────────────┘    └──────────────┘    └──────────────┘

Key Concept: The ML lifecycle never truly ends. Production models must be continuously monitored and periodically retrained as real-world data evolves.

---

4. Key ML Concepts and Terminology

4.1 Bias, Variance, and Regularization

Bias measures how far a model's predictions are from the true values. Variance measures how much predictions fluctuate across different training datasets. Every model must balance the two.

Regularization Techniques

4.2 Hyperparameters vs. Parameters

Key Concept: Parameters are learned automatically by the optimizer during training. Hyperparameters are set by the practitioner before training begins and control how learning happens.

Hyperparameter Tuning Methods

---

5. Data Fundamentals for ML

5.1 Data Types and Quality

Data Quality Dimensions

Train / Validation / Test Split

Exam Tip: Test data must remain completely unseen during training and tuning. Using test data to make any development decision invalidates the evaluation and constitutes data leakage.

5.2 Handling Missing Data and Class Imbalance

Missing Data Strategies

Class Imbalance Techniques

---

6. Feature Engineering

Feature engineering is the process of creating, transforming, or selecting input variables to improve model performance. It is often the highest-leverage activity in an ML project.

6.1 Feature Types and Transformations

Numerical Transformations

Categorical Transformations

Text Feature Methods

---

7. Model Training and Evaluation

7.1 Evaluation Metrics

Classification Metrics

The confusion matrix is the foundation for all classification metrics.

                    Predicted Positive    Predicted Negative
Actual Positive  │        TP            │        FN         │
Actual Negative  │        FP            │        TN         │

Exam Tip: For fraud detection and medical diagnosis, prioritize Recall — a missed positive (false negative) is the most costly error. For spam filtering, prioritize Precision — a false positive blocks legitimate email.

Regression Metrics

---

8. ML Algorithms Overview

8.1 Supervised Algorithms

SageMaker Built-in Algorithms

8.2 Unsupervised Algorithms

---

9. Deep Learning and Neural Networks

A neural network learns by adjusting internal weights and biases to minimize a loss function via backpropagation and gradient descent.

9.1 Network Architectures

Activation Functions

Neural Network Types

Key Concept: The Transformer architecture, introduced in "Attention Is All You Need" (2017), uses self-attention to process all tokens in parallel. It is the foundation of GPT, BERT, Claude, and every major modern language model.

Transfer Learning Steps

---

10. AWS AI and ML Services

AWS organizes its AI/ML stack into three tiers based on the level of ML expertise required.

┌──────────────────────────────────────────────────────────────┐
│  Tier 1 — AI Services  (no ML expertise needed)              │
│  Rekognition · Comprehend · Polly · Transcribe · Translate   │
├──────────────────────────────────────────────────────────────┤
│  Tier 2 — ML Services  (build and train custom models)       │
│                    Amazon SageMaker                          │
├──────────────────────────────────────────────────────────────┤
│  Tier 3 — Infrastructure  (GPUs and custom chips)            │
│        EC2 GPU instances · AWS Trainium · AWS Inferentia     │
└──────────────────────────────────────────────────────────────┘

10.1 Amazon SageMaker

Amazon SageMaker is AWS's fully managed ML platform for building, training, and deploying models at scale.

SageMaker Deployment Options

10.2 AWS Pre-Built AI Services

Vision Services

Language Services

Search, Forecasting, and Analytics Services

---

Exam Tips & Quick Reference

Scenario-to-Answer Mapping

---

Common Traps

• Accuracy on imbalanced data: Accuracy looks good on imbalanced datasets but hides poor minority-class performance. The exam will present an imbalanced scenario and expect F1, Precision, or Recall — not accuracy.
• Autopilot vs. Canvas: Autopilot automates model building for practitioners who want ML control. Canvas targets business analysts with no ML background using a no-code interface.
• Parameters vs. hyperparameters: Parameters (weights, biases) are learned automatically. Hyperparameters (learning rate, depth) are set before training. The optimizer sets parameters; the data scientist sets hyperparameters.
• Validation vs. test set: Validation data is used iteratively during development. Test data is used exactly once at the very end. Tuning on test data is data leakage.

---

Key Terms — Domain 1

---

End of Domain 1. Continue to Domain 2: Fundamentals of Generative AI →