@xdev-asia/xdev-knowledge-mcp 1.0.41 → 1.0.43

package/content/series/luyen-thi/luyen-thi-aws-ai-practitioner/chapters/01-domain-1-fundamentals-ai-ml/lessons/01-bai-1-ai-ml-deep-learning-concepts.md

@@ -0,0 +1,287 @@
+---
+id: 019c9619-lt01-d1-l01
+title: 'Bài 1: AI, ML & Deep Learning — Concepts and Terminology'
+slug: bai-1-ai-ml-deep-learning-concepts
+description: >-
+  AI vs ML vs DL. Supervised, Unsupervised, Reinforcement Learning.
+  Classification, Regression, Clustering. Neural Networks basics.
+  Training, Validation, Test sets. Bias-Variance tradeoff.
+duration_minutes: 60
+is_free: true
+video_url: null
+sort_order: 1
+section_title: "Domain 1: Fundamentals of AI and ML (20%)"
+course:
+  id: 019c9619-lt01-7001-c001-lt0100000001
+  title: 'Luyện thi AWS Certified AI Practitioner (AIF-C01)'
+  slug: luyen-thi-aws-ai-practitioner
+---
+
+<div style="text-align: center; margin: 2rem 0;">
+<img src="/storage/uploads/2026/04/aws-aif-bai1-ai-ml-dl-hierarchy.png" alt="AI, ML và Deep Learning Hierarchy" style="max-width: 800px; width: 100%; border-radius: 12px;" />
+<p><em>AI, ML và Deep Learning — quan hệ lồng nhau và ba paradigm học máy</em></p>
+</div>
+
+<h2 id="overview"><strong>Tổng quan Domain 1</strong></h2>
+
+<p>Domain 1 chiếm <strong>20% đề thi AIF-C01</strong>. Bạn cần hiểu rõ các khái niệm nền tảng về AI, ML, và Deep Learning — không cần code, nhưng phải phân biệt được khi nào dùng approach nào.</p>
+
+<blockquote>
+<p><strong>Exam tip:</strong> Domain này thường có các câu hỏi dạng "Which type of machine learning is BEST suited for..." — yêu cầu bạn chọn đúng paradigm cho use case.</p>
+</blockquote>
+
+<h2 id="ai-vs-ml-vs-dl"><strong>1. AI vs Machine Learning vs Deep Learning</strong></h2>
+
+<p>Ba khái niệm này có quan hệ lồng nhau (nested relationship):</p>
+
+<pre><code class="language-text">┌─────────────────────────────────────────────┐
+│  Artificial Intelligence (AI)               │
+│  "Machines that mimic human intelligence"   │
+│  ┌───────────────────────────────────────┐   │
+│  │  Machine Learning (ML)               │   │
+│  │  "Learning from data without         │   │
+│  │   explicit programming"              │   │
+│  │  ┌─────────────────────────────────┐  │   │
+│  │  │  Deep Learning (DL)             │  │   │
+│  │  │  "Neural networks with many     │  │   │
+│  │  │   layers"                       │  │   │
+│  │  └─────────────────────────────────┘  │   │
+│  └───────────────────────────────────────┘   │
+└─────────────────────────────────────────────┘
+</code></pre>
+
+<table>
+<thead><tr><th>Concept</th><th>Definition</th><th>Example</th></tr></thead>
+<tbody>
+<tr><td><strong>AI</strong></td><td>Broad field — machines performing tasks that typically require human intelligence</td><td>Chatbot, self-driving car, chess engine</td></tr>
+<tr><td><strong>ML</strong></td><td>Subset of AI — algorithms learn patterns from data</td><td>Spam filter, recommendation engine</td></tr>
+<tr><td><strong>DL</strong></td><td>Subset of ML — neural networks with multiple layers</td><td>Image recognition, language translation</td></tr>
+</tbody>
+</table>
+
+<h3 id="key-differences"><strong>Key Differences for the Exam</strong></h3>
+
+<ul>
+<li><strong>Traditional Programming</strong>: Rules + Data → Output</li>
+<li><strong>Machine Learning</strong>: Data + Output → Rules (model learns the rules)</li>
+<li><strong>Deep Learning</strong>: Tự động extract features từ raw data (không cần manual feature engineering)</li>
+</ul>
+
+<h2 id="ml-paradigms"><strong>2. Three ML Paradigms</strong></h2>
+
+<h3 id="supervised-learning"><strong>2.1. Supervised Learning</strong></h3>
+
+<p>Model học từ <strong>labeled data</strong> — mỗi input đi kèm output đúng (label/target).</p>
+
+<table>
+<thead><tr><th>Task Type</th><th>Output</th><th>Use Case</th><th>Algorithms</th></tr></thead>
+<tbody>
+<tr><td><strong>Classification</strong></td><td>Discrete category</td><td>Spam vs Not Spam, Fraud detection</td><td>Logistic Regression, Random Forest, SVM</td></tr>
+<tr><td><strong>Regression</strong></td><td>Continuous number</td><td>House price prediction, Stock forecast</td><td>Linear Regression, XGBoost</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> Nếu đề bài nói "predict a category" hoặc "classify" → <strong>Classification</strong>. Nếu nói "predict a number/value" → <strong>Regression</strong>.</p>
+</blockquote>
+
+<h3 id="unsupervised-learning"><strong>2.2. Unsupervised Learning</strong></h3>
+
+<p>Model học từ <strong>unlabeled data</strong> — tự tìm patterns, structure trong dữ liệu.</p>
+
+<table>
+<thead><tr><th>Task Type</th><th>What it does</th><th>Use Case</th></tr></thead>
+<tbody>
+<tr><td><strong>Clustering</strong></td><td>Group similar data points</td><td>Customer segmentation, Document grouping</td></tr>
+<tr><td><strong>Dimensionality Reduction</strong></td><td>Reduce features while preserving info</td><td>Data visualization, noise reduction</td></tr>
+<tr><td><strong>Anomaly Detection</strong></td><td>Find unusual data points</td><td>Fraud detection, equipment failure</td></tr>
+<tr><td><strong>Association</strong></td><td>Find rules between items</td><td>"Customers who bought X also bought Y"</td></tr>
+</tbody>
+</table>
+
+<h3 id="reinforcement-learning"><strong>2.3. Reinforcement Learning (RL)</strong></h3>
+
+<p>Agent học bằng cách <strong>trial-and-error</strong> trong một environment. Nhận <strong>reward</strong> (positive) hoặc <strong>penalty</strong> (negative) cho mỗi action.</p>
+
+<pre><code class="language-text">Agent → Action → Environment → State + Reward → Agent (loop)
+</code></pre>
+
+<p><strong>Use cases:</strong></p>
+<ul>
+<li>Game AI (AlphaGo)</li>
+<li>Robotics navigation</li>
+<li>Autonomous driving</li>
+<li>AWS DeepRacer (self-driving car simulation)</li>
+</ul>
+
+<h3 id="choosing-paradigm"><strong>2.4. Choosing the Right Paradigm — Exam Decision Tree</strong></h3>
+
+<pre><code class="language-text">Do you have labeled data?
+├── YES → Supervised Learning
+│   ├── Predicting a category? → Classification
+│   └── Predicting a number? → Regression
+├── NO →
+│   ├── Want to find groups/patterns? → Unsupervised (Clustering)
+│   └── Learning through trial & error? → Reinforcement Learning
+</code></pre>
+
+<h2 id="data-concepts"><strong>3. Data Concepts for ML</strong></h2>
+
+<h3 id="data-types"><strong>3.1. Data Types</strong></h3>
+
+<table>
+<thead><tr><th>Type</th><th>Description</th><th>Examples</th></tr></thead>
+<tbody>
+<tr><td><strong>Structured</strong></td><td>Organized in rows & columns (tabular)</td><td>CSV, database tables, spreadsheets</td></tr>
+<tr><td><strong>Semi-structured</strong></td><td>Has some organization but flexible</td><td>JSON, XML, log files</td></tr>
+<tr><td><strong>Unstructured</strong></td><td>No predefined format</td><td>Images, videos, audio, free text</td></tr>
+<tr><td><strong>Time-series</strong></td><td>Data points indexed by time</td><td>Stock prices, IoT sensor readings</td></tr>
+</tbody>
+</table>
+
+<h3 id="labeled-unlabeled"><strong>3.2. Labeled vs Unlabeled Data</strong></h3>
+
+<ul>
+<li><strong>Labeled data</strong>: Mỗi data point có kèm answer (label). Ví dụ: email + tag "spam"/"not spam". Dùng cho <strong>Supervised Learning</strong>.</li>
+<li><strong>Unlabeled data</strong>: Chỉ có data, không có label. Dùng cho <strong>Unsupervised Learning</strong>.</li>
+<li><strong>Amazon SageMaker Ground Truth</strong>: Dịch vụ AWS giúp label data (human + ML-assisted labeling).</li>
+</ul>
+
+<h3 id="datasets"><strong>3.3. Training, Validation, Test Sets</strong></h3>
+
+<pre><code class="language-text">┌────────────────────────────────────────────────┐
+│              Full Dataset (100%)               │
+├──────────────────┬──────────┬──────────────────┤
+│  Training (70%)  │ Val(15%) │   Test (15%)     │
+│  Model learns    │ Tune     │ Final evaluation │
+│  from this data  │ hyper-   │ (never seen      │
+│                  │ params   │  during training) │
+└──────────────────┴──────────┴──────────────────┘
+</code></pre>
+
+<ul>
+<li><strong>Training set</strong>: Model học patterns từ đây</li>
+<li><strong>Validation set</strong>: Tune hyperparameters, chống overfitting</li>
+<li><strong>Test set</strong>: Đánh giá cuối cùng — model chưa bao giờ thấy data này</li>
+</ul>
+
+<h2 id="neural-networks"><strong>4. Neural Networks Basics</strong></h2>
+
+<h3 id="nn-architecture"><strong>4.1. Architecture</strong></h3>
+
+<pre><code class="language-text">Input Layer → Hidden Layer(s) → Output Layer
+    x₁ ──┐     ┌── h₁ ──┐
+    x₂ ──┼─────┼── h₂ ──┼──── ŷ (prediction)
+    x₃ ──┘     └── h₃ ──┘
+
+Each connection has a weight (w)
+Each neuron applies an activation function
+</code></pre>
+
+<p><strong>Key components:</strong></p>
+<ul>
+<li><strong>Weights</strong>: Parameters the model learns during training</li>
+<li><strong>Bias</strong>: Additional parameter to shift the activation function</li>
+<li><strong>Activation Function</strong>: ReLU, Sigmoid, Softmax — introduces non-linearity</li>
+<li><strong>Loss Function</strong>: Measures how wrong the model's predictions are</li>
+<li><strong>Optimizer</strong>: Updates weights to minimize loss (e.g., SGD, Adam)</li>
+</ul>
+
+<h3 id="nn-types"><strong>4.2. Types of Neural Networks</strong></h3>
+
+<table>
+<thead><tr><th>Type</th><th>Best For</th><th>AWS Service</th></tr></thead>
+<tbody>
+<tr><td><strong>CNN</strong> (Convolutional NN)</td><td>Images, video</td><td>Amazon Rekognition</td></tr>
+<tr><td><strong>RNN/LSTM</strong> (Recurrent NN)</td><td>Sequential data, time series</td><td>Amazon Forecast</td></tr>
+<tr><td><strong>Transformer</strong></td><td>NLP, text generation</td><td>Amazon Bedrock (LLMs)</td></tr>
+<tr><td><strong>GAN</strong> (Generative Adversarial)</td><td>Generate new data (images)</td><td>—</td></tr>
+</tbody>
+</table>
+
+<h2 id="model-evaluation"><strong>5. Model Evaluation Concepts</strong></h2>
+
+<h3 id="overfitting-underfitting"><strong>5.1. Overfitting vs Underfitting</strong></h3>
+
+<table>
+<thead><tr><th>Problem</th><th>Training Accuracy</th><th>Test Accuracy</th><th>Cause</th><th>Solution</th></tr></thead>
+<tbody>
+<tr><td><strong>Overfitting</strong></td><td>Very High</td><td>Low</td><td>Model memorizes training data</td><td>More data, regularization, dropout, early stopping</td></tr>
+<tr><td><strong>Underfitting</strong></td><td>Low</td><td>Low</td><td>Model too simple</td><td>More features, more complex model, longer training</td></tr>
+<tr><td><strong>Good Fit</strong></td><td>High</td><td>High</td><td>Balanced complexity</td><td>—</td></tr>
+</tbody>
+</table>
+
+<h3 id="bias-variance"><strong>5.2. Bias-Variance Tradeoff</strong></h3>
+
+<ul>
+<li><strong>High Bias</strong> = Underfitting (model quá đơn giản, bỏ qua patterns)</li>
+<li><strong>High Variance</strong> = Overfitting (model quá phức tạp, nhạy cảm với noise)</li>
+<li>Mục tiêu: tìm <strong>sweet spot</strong> giữa bias và variance</li>
+</ul>
+
+<h3 id="metrics"><strong>5.3. Common Metrics</strong></h3>
+
+<p><strong>Classification metrics:</strong></p>
+<table>
+<thead><tr><th>Metric</th><th>Formula</th><th>When to use</th></tr></thead>
+<tbody>
+<tr><td><strong>Accuracy</strong></td><td>(TP + TN) / Total</td><td>Balanced classes</td></tr>
+<tr><td><strong>Precision</strong></td><td>TP / (TP + FP)</td><td>"Don't flag innocent as spam"</td></tr>
+<tr><td><strong>Recall</strong></td><td>TP / (TP + FN)</td><td>"Don't miss any fraud"</td></tr>
+<tr><td><strong>F1 Score</strong></td><td>2 × (P × R) / (P + R)</td><td>Imbalanced classes</td></tr>
+<tr><td><strong>AUC-ROC</strong></td><td>Area under ROC curve</td><td>Binary classification overall</td></tr>
+</tbody>
+</table>
+
+<p><strong>Regression metrics:</strong></p>
+<ul>
+<li><strong>RMSE</strong> (Root Mean Square Error): Penalizes large errors</li>
+<li><strong>MAE</strong> (Mean Absolute Error): Average error magnitude</li>
+<li><strong>R²</strong>: How well model explains variance (1.0 = perfect)</li>
+</ul>
+
+<h2 id="key-terms"><strong>6. Key Terms Cheat Sheet</strong></h2>
+
+<table>
+<thead><tr><th>Term</th><th>Definition (for exam)</th></tr></thead>
+<tbody>
+<tr><td><strong>Feature</strong></td><td>Input variable used for prediction (column in data)</td></tr>
+<tr><td><strong>Label / Target</strong></td><td>The answer we want the model to predict</td></tr>
+<tr><td><strong>Hyperparameter</strong></td><td>Settings configured BEFORE training (learning rate, epochs)</td></tr>
+<tr><td><strong>Parameter</strong></td><td>Values the model learns DURING training (weights, biases)</td></tr>
+<tr><td><strong>Epoch</strong></td><td>One complete pass through the entire training dataset</td></tr>
+<tr><td><strong>Batch Size</strong></td><td>Number of samples processed before updating weights</td></tr>
+<tr><td><strong>Inference</strong></td><td>Using a trained model to make predictions on new data</td></tr>
+<tr><td><strong>Transfer Learning</strong></td><td>Using a pre-trained model and adapting it for a new task</td></tr>
+</tbody>
+</table>
+
+<h2 id="practice-questions"><strong>7. Practice Questions</strong></h2>
+
+<p><strong>Q1:</strong> A company wants to predict whether customers will cancel their subscription (yes/no). Which ML approach is most appropriate?</p>
+<ul>
+<li>A) Unsupervised Learning — Clustering</li>
+<li>B) Supervised Learning — Regression</li>
+<li>C) Supervised Learning — Classification ✓</li>
+<li>D) Reinforcement Learning</li>
+</ul>
+<p><em>Explanation: Predicting a binary outcome (yes/no) with labeled historical data = supervised classification.</em></p>
+
+<p><strong>Q2:</strong> A retail company has customer purchase data but NO predefined groups. They want to segment customers into groups for targeted marketing. Which approach should they use?</p>
+<ul>
+<li>A) Supervised Learning — Classification</li>
+<li>B) Unsupervised Learning — Clustering ✓</li>
+<li>C) Reinforcement Learning</li>
+<li>D) Supervised Learning — Regression</li>
+</ul>
+<p><em>Explanation: No labels + finding natural groups in data = unsupervised clustering.</em></p>
+
+<p><strong>Q3:</strong> A model performs extremely well on training data (99% accuracy) but poorly on new data (65% accuracy). What is this called?</p>
+<ul>
+<li>A) Underfitting</li>
+<li>B) Overfitting ✓</li>
+<li>C) High bias</li>
+<li>D) Regularization</li>
+</ul>
+<p><em>Explanation: High training accuracy + low test accuracy = overfitting (model memorized training data).</em></p>

package/content/series/luyen-thi/luyen-thi-aws-ai-practitioner/chapters/01-domain-1-fundamentals-ai-ml/lessons/02-bai-2-ml-lifecycle-aws-services.md

@@ -0,0 +1,258 @@
+---
+id: 019c9619-lt01-d1-l02
+title: 'Bài 2: ML Development Lifecycle & AWS AI Services Overview'
+slug: bai-2-ml-lifecycle-aws-services
+description: >-
+  ML pipeline: data collection → feature engineering → training → evaluation → deployment.
+  AWS AI/ML service stack. SageMaker, Rekognition, Comprehend, Polly,
+  Transcribe, Translate, Textract, Lex, Personalize, Forecast, Kendra.
+duration_minutes: 60
+is_free: true
+video_url: null
+sort_order: 2
+section_title: "Domain 1: Fundamentals of AI and ML (20%)"
+course:
+  id: 019c9619-lt01-7001-c001-lt0100000001
+  title: 'Luyện thi AWS Certified AI Practitioner (AIF-C01)'
+  slug: luyen-thi-aws-ai-practitioner
+---
+
+<div style="text-align: center; margin: 2rem 0;">
+<img src="/storage/uploads/2026/04/aws-aif-bai2-ml-lifecycle-pipeline.png" alt="ML Development Lifecycle Pipeline on AWS" style="max-width: 800px; width: 100%; border-radius: 12px;" />
+<p><em>ML Development Lifecycle Pipeline và AWS AI/ML Service Stack</em></p>
+</div>
+
+<h2 id="ml-lifecycle"><strong>1. ML Development Lifecycle</strong></h2>
+
+<p>Đề thi AIF-C01 yêu cầu bạn hiểu toàn bộ vòng đời phát triển ML — từ khi xác định bài toán đến khi deploy và monitor model.</p>
+
+<pre><code class="language-text">┌─────────────┐    ┌──────────────┐    ┌──────────────┐
+│ 1. Business │───→│ 2. Data      │───→│ 3. Feature   │
+│ Problem     │    │ Collection & │    │ Engineering  │
+│ Definition  │    │ Preparation  │    │              │
+└─────────────┘    └──────────────┘    └──────────────┘
+                                              │
+┌─────────────┐    ┌──────────────┐    ┌──────┴───────┐
+│ 6. Monitor  │←───│ 5. Deploy    │←───│ 4. Model     │
+│ & Retrain   │    │ & Inference  │    │ Training &   │
+│             │    │              │    │ Evaluation   │
+└─────────────┘    └──────────────┘    └──────────────┘
+</code></pre>
+
+<h3 id="step-1"><strong>Step 1: Business Problem Definition</strong></h3>
+
+<ul>
+<li>Xác định bài toán có thực sự cần ML không (sometimes rules-based is enough)</li>
+<li>Define success metrics (KPIs)</li>
+<li>Determine data availability</li>
+</ul>
+
+<blockquote>
+<p><strong>Exam tip:</strong> "Not every problem needs ML." Nếu đề bài mô tả bài toán đơn giản, có thể rule-based hoặc lookup table là đủ.</p>
+</blockquote>
+
+<h3 id="step-2"><strong>Step 2: Data Collection & Preparation</strong></h3>
+
+<ul>
+<li><strong>Data Collection</strong>: Thu thập từ databases, APIs, IoT, logs</li>
+<li><strong>Data Cleaning</strong>: Handle missing values, remove duplicates, fix errors</li>
+<li><strong>Data Labeling</strong>: Gắn nhãn cho supervised learning → <strong>Amazon SageMaker Ground Truth</strong></li>
+<li><strong>Exploratory Data Analysis (EDA)</strong>: Visualize, understand distributions, correlations</li>
+</ul>
+
+<h3 id="step-3"><strong>Step 3: Feature Engineering</strong></h3>
+
+<ul>
+<li><strong>Feature selection</strong>: Chọn features quan trọng, loại bỏ noise</li>
+<li><strong>Feature transformation</strong>: Normalization, scaling, encoding</li>
+<li><strong>Feature creation</strong>: Tạo features mới từ raw data</li>
+<li>AWS: <strong>SageMaker Data Wrangler</strong>, <strong>SageMaker Feature Store</strong></li>
+</ul>
+
+<h3 id="step-4"><strong>Step 4: Model Training & Evaluation</strong></h3>
+
+<ul>
+<li>Choose algorithm appropriate for the problem</li>
+<li>Split data into training/validation/test sets</li>
+<li>Train model, tune hyperparameters</li>
+<li>Evaluate using appropriate metrics (accuracy, F1, RMSE...)</li>
+<li>AWS: <strong>Amazon SageMaker</strong> for full ML workflow</li>
+</ul>
+
+<h3 id="step-5"><strong>Step 5: Deployment & Inference</strong></h3>
+
+<ul>
+<li><strong>Real-time inference</strong>: Endpoint cho instant predictions</li>
+<li><strong>Batch inference</strong>: Process large datasets offline</li>
+<li><strong>Edge deployment</strong>: Run model on edge devices</li>
+<li>AWS: <strong>SageMaker Endpoints</strong>, <strong>Lambda</strong>, <strong>IoT Greengrass</strong></li>
+</ul>
+
+<h3 id="step-6"><strong>Step 6: Monitoring & Retraining</strong></h3>
+
+<ul>
+<li><strong>Model drift</strong>: Performance degrades over time as data changes</li>
+<li><strong>Data drift</strong>: Input data distribution changes</li>
+<li><strong>Concept drift</strong>: Relationship between input and output changes</li>
+<li>Solution: Monitor → detect drift → retrain with new data</li>
+<li>AWS: <strong>SageMaker Model Monitor</strong></li>
+</ul>
+
+<h2 id="aws-ai-stack"><strong>2. AWS AI/ML Service Stack</strong></h2>
+
+<p>AWS cung cấp 3 layers of AI/ML services — từ high-level (no ML knowledge needed) đến low-level (full control):</p>
+
+<pre><code class="language-text">┌─────────────────────────────────────────────────────┐
+│  Layer 3: AI Services (Pre-trained, API-based)      │
+│  → Rekognition, Comprehend, Polly, Transcribe,      │
+│    Translate, Textract, Lex, Personalize, Forecast   │
+│  → NO ML expertise needed                           │
+├─────────────────────────────────────────────────────┤
+│  Layer 2: ML Services (Managed platform)            │
+│  → Amazon SageMaker, SageMaker JumpStart            │
+│  → Amazon Bedrock (GenAI)                           │
+│  → SOME ML expertise needed                         │
+├─────────────────────────────────────────────────────┤
+│  Layer 1: ML Frameworks & Infrastructure            │
+│  → EC2 with GPU/Inferentia, Deep Learning AMIs,     │
+│    Deep Learning Containers                         │
+│  → FULL ML expertise needed                         │
+└─────────────────────────────────────────────────────┘
+</code></pre>
+
+<h2 id="ai-services"><strong>3. AWS AI Services — Bảng Tổng hợp</strong></h2>
+
+<p>Đây là phần <strong>rất quan trọng cho đề thi</strong> — bạn cần biết mỗi service làm gì và khi nào dùng.</p>
+
+<h3 id="vision"><strong>3.1. Computer Vision</strong></h3>
+
+<table>
+<thead><tr><th>Service</th><th>What it does</th><th>Use Cases</th></tr></thead>
+<tbody>
+<tr><td><strong>Amazon Rekognition</strong></td><td>Image and video analysis</td><td>Face detection, object detection, content moderation, celebrity recognition, text in images (OCR)</td></tr>
+<tr><td><strong>Amazon Textract</strong></td><td>Extract text & data from documents</td><td>Invoice processing, ID document extraction, form data, table extraction</td></tr>
+<tr><td><strong>Amazon Lookout for Vision</strong></td><td>Visual inspection for manufacturing</td><td>Defect detection in products on assembly line</td></tr>
+</tbody>
+</table>
+
+<h3 id="nlp"><strong>3.2. Natural Language Processing (NLP)</strong></h3>
+
+<table>
+<thead><tr><th>Service</th><th>What it does</th><th>Use Cases</th></tr></thead>
+<tbody>
+<tr><td><strong>Amazon Comprehend</strong></td><td>NLP analysis</td><td>Sentiment analysis, entity extraction, key phrases, language detection, PII detection</td></tr>
+<tr><td><strong>Amazon Translate</strong></td><td>Neural machine translation</td><td>Real-time translation, batch document translation</td></tr>
+<tr><td><strong>Amazon Kendra</strong></td><td>Intelligent enterprise search</td><td>Internal knowledge search, FAQ, document search powered by NLP</td></tr>
+</tbody>
+</table>
+
+<h3 id="speech"><strong>3.3. Speech</strong></h3>
+
+<table>
+<thead><tr><th>Service</th><th>What it does</th><th>Direction</th></tr></thead>
+<tbody>
+<tr><td><strong>Amazon Polly</strong></td><td>Text-to-Speech (TTS)</td><td>Text → Audio</td></tr>
+<tr><td><strong>Amazon Transcribe</strong></td><td>Speech-to-Text (STT)</td><td>Audio → Text</td></tr>
+<tr><td><strong>Amazon Lex</strong></td><td>Conversational AI (chatbot)</td><td>Build chatbots with voice & text (powers Alexa)</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> Polly = text TO speech (Polly "speaks"). Transcribe = speech TO text (Transcribe "writes down").</p>
+</blockquote>
+
+<h3 id="predictions"><strong>3.4. Predictions & Recommendations</strong></h3>
+
+<table>
+<thead><tr><th>Service</th><th>What it does</th><th>Use Cases</th></tr></thead>
+<tbody>
+<tr><td><strong>Amazon Personalize</strong></td><td>Real-time personalization & recommendations</td><td>Product recommendations, personalized content</td></tr>
+<tr><td><strong>Amazon Forecast</strong></td><td>Time-series forecasting</td><td>Demand planning, financial forecasting, resource planning</td></tr>
+<tr><td><strong>Amazon Fraud Detector</strong></td><td>Detect online fraud</td><td>Payment fraud, fake accounts, account takeover</td></tr>
+</tbody>
+</table>
+
+<h2 id="sagemaker"><strong>4. Amazon SageMaker Overview</strong></h2>
+
+<p>SageMaker là <strong>fully managed ML platform</strong> — cung cấp mọi thứ cần thiết cho toàn bộ ML lifecycle.</p>
+
+<h3 id="sagemaker-components"><strong>Key Components:</strong></h3>
+
+<table>
+<thead><tr><th>Component</th><th>Purpose</th></tr></thead>
+<tbody>
+<tr><td><strong>SageMaker Studio</strong></td><td>IDE cho ML development (Jupyter-based)</td></tr>
+<tr><td><strong>SageMaker Ground Truth</strong></td><td>Data labeling service (human + ML-assisted)</td></tr>
+<tr><td><strong>SageMaker Data Wrangler</strong></td><td>Data preparation & transformation (no code)</td></tr>
+<tr><td><strong>SageMaker Feature Store</strong></td><td>Store & share ML features</td></tr>
+<tr><td><strong>SageMaker Training</strong></td><td>Managed training jobs with built-in algorithms</td></tr>
+<tr><td><strong>SageMaker Autopilot</strong></td><td>AutoML — automatic model building</td></tr>
+<tr><td><strong>SageMaker JumpStart</strong></td><td>Pre-trained models & solutions (model hub)</td></tr>
+<tr><td><strong>SageMaker Endpoints</strong></td><td>Deploy models for real-time inference</td></tr>
+<tr><td><strong>SageMaker Model Monitor</strong></td><td>Monitor deployed models for drift</td></tr>
+<tr><td><strong>SageMaker Clarify</strong></td><td>Bias detection & model explainability</td></tr>
+<tr><td><strong>SageMaker Canvas</strong></td><td>No-code ML for business users</td></tr>
+</tbody>
+</table>
+
+<h3 id="sagemaker-decision"><strong>When to use SageMaker vs AI Services?</strong></h3>
+
+<pre><code class="language-text">Need custom ML model? → SageMaker
+Need pre-trained capability? → AI Services (Rekognition, Comprehend, etc.)
+Need GenAI/Foundation Models? → Amazon Bedrock
+Business user, no code? → SageMaker Canvas
+</code></pre>
+
+<h2 id="service-mapping"><strong>5. Use Case → AWS Service Mapping</strong></h2>
+
+<p>Đây là dạng câu hỏi rất phổ biến trong đề thi:</p>
+
+<table>
+<thead><tr><th>Use Case</th><th>AWS Service</th></tr></thead>
+<tbody>
+<tr><td>Detect faces in photos</td><td>Amazon Rekognition</td></tr>
+<tr><td>Extract data from invoices</td><td>Amazon Textract</td></tr>
+<tr><td>Analyze customer review sentiment</td><td>Amazon Comprehend</td></tr>
+<tr><td>Translate content to multiple languages</td><td>Amazon Translate</td></tr>
+<tr><td>Build a customer service chatbot</td><td>Amazon Lex</td></tr>
+<tr><td>Convert blog posts to audio</td><td>Amazon Polly</td></tr>
+<tr><td>Transcribe meeting recordings</td><td>Amazon Transcribe</td></tr>
+<tr><td>Product recommendations</td><td>Amazon Personalize</td></tr>
+<tr><td>Demand forecasting</td><td>Amazon Forecast</td></tr>
+<tr><td>Search internal documents</td><td>Amazon Kendra</td></tr>
+<tr><td>Detect fraudulent transactions</td><td>Amazon Fraud Detector</td></tr>
+<tr><td>Label training data</td><td>SageMaker Ground Truth</td></tr>
+<tr><td>Build custom ML model</td><td>Amazon SageMaker</td></tr>
+<tr><td>No-code ML for business analysts</td><td>SageMaker Canvas</td></tr>
+<tr><td>Generate text with LLM</td><td>Amazon Bedrock</td></tr>
+</tbody>
+</table>
+
+<h2 id="practice-questions"><strong>6. Practice Questions</strong></h2>
+
+<p><strong>Q1:</strong> A company wants to automatically extract text and structured data from scanned invoices. Which AWS service should they use?</p>
+<ul>
+<li>A) Amazon Comprehend</li>
+<li>B) Amazon Rekognition</li>
+<li>C) Amazon Textract ✓</li>
+<li>D) Amazon Translate</li>
+</ul>
+<p><em>Explanation: Textract is specifically designed to extract text, forms, and tables from scanned documents. Comprehend analyzes text meaning, not document extraction. Rekognition is for image/video analysis.</em></p>
+
+<p><strong>Q2:</strong> A data scientist notices that their deployed model's prediction accuracy has decreased over the past month. The input data patterns have changed. What is this called?</p>
+<ul>
+<li>A) Overfitting</li>
+<li>B) Underfitting</li>
+<li>C) Data drift ✓</li>
+<li>D) Feature engineering</li>
+</ul>
+<p><em>Explanation: When the statistical properties of model input data change over time, causing performance degradation, this is called data drift.</em></p>
+
+<p><strong>Q3:</strong> Which AWS service allows business analysts with no ML experience to build ML models using a visual interface?</p>
+<ul>
+<li>A) SageMaker Studio</li>
+<li>B) SageMaker Autopilot</li>
+<li>C) SageMaker Canvas ✓</li>
+<li>D) SageMaker JumpStart</li>
+</ul>
+<p><em>Explanation: SageMaker Canvas provides a no-code, visual point-and-click interface for business analysts. Autopilot automates model building but requires some ML knowledge. JumpStart provides pre-trained models. Studio is the full ML IDE.</em></p>