@xdev-asia/xdev-knowledge-mcp 1.0.43 → 1.0.45

package/content/series/luyen-thi/luyen-thi-aws-ml-specialty/chapters/03-phan-3-implementation-operations/lessons/09-bai-9-security-cost.md

@@ -0,0 +1,166 @@
+---
+id: bb3d4aa7-2e63-49f6-a751-6323c5919325
+title: 'Bài 9: Security & Cost Optimization'
+slug: bai-9-security-cost
+description: >-
+  IAM roles và policies cho SageMaker. VPC configuration, PrivateLink.
+  Encryption at rest (KMS) và in transit. Spot Training Instances.
+  S3 lifecycle policies cho ML data. Right-sizing instances.
+duration_minutes: 45
+is_free: true
+video_url: null
+sort_order: 9
+section_title: "Phần 3: ML Implementation & Operations (20%)"
+course:
+  id: 019c9619-lt02-7002-c002-lt0200000002
+  title: 'Luyện thi AWS Certified Machine Learning - Specialty'
+  slug: luyen-thi-aws-ml-specialty
+---
+
+<div style="text-align: center; margin: 2rem 0;">
+<img src="/storage/uploads/2026/04/aws-mls-bai9-security-architecture.png" alt="AWS ML Security Architecture" style="max-width: 800px; width: 100%; border-radius: 12px;" />
+<p><em>Security trong AWS ML: IAM Roles, VPC isolation, KMS encryption, và tối ưu chi phí với Spot Instances</em></p>
+</div>
+
+<h2 id="iam-sagemaker"><strong>1. IAM for SageMaker</strong></h2>
+
+<p>SageMaker sử dụng <strong>IAM Roles</strong> (không phải users) để thực hiện actions trên AWS resources. Đây là pattern bảo mật quan trọng trong đề thi.</p>
+
+<table>
+<thead><tr><th>Role Type</th><th>Used By</th><th>Needs Access To</th></tr></thead>
+<tbody>
+<tr><td><strong>Execution Role</strong></td><td>SageMaker Notebooks, Training Jobs, Endpoints</td><td>S3, ECR, CloudWatch, KMS</td></tr>
+<tr><td><strong>SageMaker Studio Role</strong></td><td>Studio IDE users</td><td>Data, experiments, pipelines</td></tr>
+<tr><td><strong>Training Job Role</strong></td><td>The training container itself</td><td>Input/output S3 buckets</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> SageMaker training/inference containers KHÔNG có EC2 instance credentials — họ chạy với IAM Role cross-account. Luôn cần grant S3 và ECR permissions cho execution role.</p>
+</blockquote>
+
+<h2 id="vpc-security"><strong>2. VPC Configuration for SageMaker</strong></h2>
+
+<p>Chạy SageMaker workloads trong <strong>VPC</strong> để đảm bảo traffic không đi qua public internet.</p>
+
+<pre><code class="language-text">SageMaker Network Security:
+
+Internet ──✗────────────────────────────────────────
+                                                    │
+          ┌─── Private VPC ──────────────────────┐ │
+          │                                       │ │
+          │  SageMaker Training Instance          │ │
+          │          ↓ (VPC Endpoint)             │ │
+          │  ┌──── S3 Gateway Endpoint ────────┐  │ │
+          │  │   ECR VPC Endpoint              │  │ │
+          │  │   SageMaker API VPC Endpoint    │  │ │
+          │  └─────────────────────────────────┘  │ │
+          └───────────────────────────────────────┘ │
+</code></pre>
+
+<table>
+<thead><tr><th>Feature</th><th>Description</th></tr></thead>
+<tbody>
+<tr><td><strong>VPC Endpoints (PrivateLink)</strong></td><td>Access S3, ECR, SageMaker API without internet</td></tr>
+<tr><td><strong>Security Groups</strong></td><td>Control inbound/outbound traffic cho training instances</td></tr>
+<tr><td><strong>Network Isolation</strong></td><td>Training job không có internet access (isolated mode)</td></tr>
+<tr><td><strong>Inter-Container Encryption</strong></td><td>Encrypt distributed training traffic</td></tr>
+</tbody>
+</table>
+
+<h2 id="encryption"><strong>3. Encryption</strong></h2>
+
+<table>
+<thead><tr><th>What</th><th>How</th><th>Service</th></tr></thead>
+<tbody>
+<tr><td><strong>S3 data at rest</strong></td><td>SSE-S3, SSE-KMS, SSE-C</td><td>S3 + KMS</td></tr>
+<tr><td><strong>Model artifacts at rest</strong></td><td>KMS key cho output S3 bucket</td><td>KMS</td></tr>
+<tr><td><strong>EBS volumes (training)</strong></td><td>KMS encryption for instance storage</td><td>KMS</td></tr>
+<tr><td><strong>Data in transit</strong></td><td>TLS 1.2/1.3 for all API calls</td><td>Default</td></tr>
+<tr><td><strong>Distributed training traffic</strong></td><td>Enable inter-container encryption</td><td>SageMaker config</td></tr>
+</tbody>
+</table>
+
+<h2 id="cost-optimization"><strong>4. Cost Optimization Strategies</strong></h2>
+
+<table>
+<thead><tr><th>Strategy</th><th>Savings</th><th>How</th></tr></thead>
+<tbody>
+<tr><td><strong>Spot Instances</strong></td><td>Up to 90%</td><td>Training Jobs + checkpointing</td></tr>
+<tr><td><strong>Right-sizing</strong></td><td>20-40%</td><td>Match instance type to actual GPU/CPU usage</td></tr>
+<tr><td><strong>Serverless Inference</strong></td><td>Variable</td><td>Pay per invocation, no idle cost</td></tr>
+<tr><td><strong>SageMaker Savings Plans</strong></td><td>Up to 64%</td><td>Commit to consistent usage</td></tr>
+<tr><td><strong>S3 Intelligent-Tiering</strong></td><td>Variable</td><td>Auto-tier old training data</td></tr>
+<tr><td><strong>Lifecycle Configurations</strong></td><td>Variable</td><td>Auto-stop idle notebooks</td></tr>
+</tbody>
+</table>
+
+<h3 id="s3-lifecycle"><strong>4.1. S3 Lifecycle Policies cho ML Data</strong></h3>
+
+<pre><code class="language-text">Data Lifecycle for ML:
+
+  Active Training Data (S3 Standard)
+           ↓ after 30 days unused
+  S3 Intelligent-Tiering
+           ↓ after 90 days
+  S3 Standard-IA (Infrequent Access)
+           ↓ after 180 days
+  S3 Glacier Instant Retrieval
+           ↓ after 1 year
+  S3 Glacier Deep Archive (compliance)
+</code></pre>
+
+<h2 id="compliance"><strong>5. Compliance Frameworks</strong></h2>
+
+<table>
+<thead><tr><th>Framework</th><th>Relevance for ML</th></tr></thead>
+<tbody>
+<tr><td><strong>HIPAA</strong></td><td>Healthcare ML — PHI data encryption, audit logging, BAA required</td></tr>
+<tr><td><strong>GDPR</strong></td><td>EU data — right to erasure, data minimization, consent</td></tr>
+<tr><td><strong>SOC 2</strong></td><td>Security controls audit for SaaS ML products</td></tr>
+<tr><td><strong>PCI DSS</strong></td><td>Payment card data in ML models</td></tr>
+</tbody>
+</table>
+
+<h2 id="cheat-sheet"><strong>6. Cheat Sheet — Security & Cost</strong></h2>
+
+<table>
+<thead><tr><th>Scenario</th><th>Solution</th></tr></thead>
+<tbody>
+<tr><td>SageMaker training with no internet</td><td>VPC + Network Isolation + VPC Endpoints</td></tr>
+<tr><td>Encrypt training data on S3</td><td>SSE-KMS with customer-managed key</td></tr>
+<tr><td>Reduce training cost by 70%+</td><td>Spot Instances + checkpointing</td></tr>
+<tr><td>Auto-archive old training datasets</td><td>S3 Lifecycle Policies</td></tr>
+<tr><td>Prevent notebook idle cost</td><td>Studio Lifecycle Config → auto-shutdown</td></tr>
+<tr><td>Healthcare data (HIPAA)</td><td>KMS + VPC + CloudTrail + BAA with AWS</td></tr>
+</tbody>
+</table>
+
+<h2 id="practice"><strong>7. Practice Questions</strong></h2>
+
+<p><strong>Q1:</strong> A company needs SageMaker training jobs to access data in S3 without traversing the public internet for security compliance. What should they configure?</p>
+<ul>
+<li>A) VPC Flow Logs</li>
+<li>B) SageMaker Training with VPC + S3 VPC Gateway Endpoint ✓</li>
+<li>C) IAM policy with IP restriction</li>
+<li>D) AWS Shield</li>
+</ul>
+<p><em>Explanation: Configuring SageMaker Training Jobs to run in a VPC, combined with an S3 VPC Gateway Endpoint, ensures all S3 traffic stays within the AWS network without going through the public internet.</em></p>
+
+<p><strong>Q2:</strong> A machine learning team wants to reduce costs for long-running training jobs that can be interrupted. The jobs should resume from where they stopped. Which approach is MOST cost-effective?</p>
+<ul>
+<li>A) Use larger instances to finish faster</li>
+<li>B) Use Reserved Instances</li>
+<li>C) Use Spot Instances with checkpointing to S3 ✓</li>
+<li>D) Run training locally</li>
+</ul>
+<p><em>Explanation: Spot Instances provide up to 90% cost savings. With checkpointing enabled (saving model state to S3 periodically), jobs can resume from the last checkpoint if interrupted, making Spot Instances practical for long training runs.</em></p>
+
+<p><strong>Q3:</strong> Which AWS service provides centralized key management for encrypting SageMaker training data, model artifacts, and EBS volumes?</p>
+<ul>
+<li>A) AWS Secrets Manager</li>
+<li>B) AWS IAM</li>
+<li>C) AWS KMS (Key Management Service) ✓</li>
+<li>D) AWS Certificate Manager</li>
+</ul>
+<p><em>Explanation: AWS KMS provides encryption key management for at-rest encryption of S3 data (SSE-KMS), EBS volumes used by training instances, and model artifacts. SageMaker integrates natively with KMS throughout the training and deployment workflow.</em></p>

package/content/series/luyen-thi/luyen-thi-aws-ml-specialty/chapters/04-phan-4-on-tap/lessons/10-bai-10-bai-toan-thuong-gap.md

@@ -0,0 +1,181 @@
+---
+id: f7a8daa0-e49a-4fac-b232-d6a8b998e120
+title: 'Bài 10: Các Bài Toán ML Thường Gặp'
+slug: bai-10-bai-toan-thuong-gap
+description: >-
+  Fraud detection, recommendation systems, NLP pipeline, time series forecasting,
+  computer vision — nhận dạng bài toán và chọn đúng AWS service/algorithm.
+duration_minutes: 55
+is_free: true
+video_url: null
+sort_order: 10
+section_title: "Phần 4: Ôn Tập Tổng Hợp"
+course:
+  id: 019c9619-lt02-7002-c002-lt0200000002
+  title: 'Luyện thi AWS Certified Machine Learning - Specialty'
+  slug: luyen-thi-aws-ml-specialty
+---
+
+<div style="text-align: center; margin: 2rem 0;">
+<img src="/storage/uploads/2026/04/aws-mls-bai10-ml-patterns.png" alt="AWS ML Problem Patterns" style="max-width: 800px; width: 100%; border-radius: 12px;" />
+<p><em>ML Problem Patterns: Fraud detection, Recommendation, NLP, Time Series, và Computer Vision trên AWS</em></p>
+</div>
+
+<h2 id="pattern-matching"><strong>1. Nhận Dạng Bài Toán Nhanh</strong></h2>
+
+<p>Phần lớn câu hỏi MLS-C01 là <strong>scenario-based</strong>: đọc mô tả business problem, chọn AWS service hoặc algorithm phù hợp nhất. Bài học này cung cấp framework để nhận dạng nhanh.</p>
+
+<pre><code class="language-text">ML Problem Recognition Framework:
+
+READ SCENARIO
+    ↓
+Is output a CATEGORY?          → Classification
+Is output a NUMBER?            → Regression
+Is output a CLUSTER/GROUP?     → Clustering (no labels)
+Is output a RANK/SCORE?        → Recommendation / Ranking
+Is output a DETECTION (anomaly)? → Anomaly Detection
+Is output a SEQUENCE?          → Time Series / Seq2Seq
+</code></pre>
+
+<h2 id="fraud-detection"><strong>2. Fraud Detection</strong></h2>
+
+<p>Fraud detection là bài toán điển hình trong đề thi. Key challenge: <strong>extreme class imbalance</strong> (99.9% transactions là legitimate).</p>
+
+<table>
+<thead><tr><th>Challenge</th><th>Solution</th></tr></thead>
+<tbody>
+<tr><td>Class imbalance</td><td>SMOTE, class_weight, <strong>precision-recall</strong> (not accuracy)</td></tr>
+<tr><td>Real-time scoring</td><td>SageMaker Real-time Endpoint</td></tr>
+<tr><td>Unsupervised fraud (no labels)</td><td><strong>Random Cut Forest</strong> (anomaly detection)</td></tr>
+<tr><td>Graph-based fraud rings</td><td><strong>Amazon Neptune</strong> + GNN (GraphSAGE)</td></tr>
+<tr><td>Supervised (labeled history)</td><td><strong>XGBoost</strong> or Linear Learner</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> Khi câu hỏi nói "không có nhãn fraud" hoặc "phát hiện hành vi bất thường" → <strong>Random Cut Forest</strong>. Khi có lịch sử giao dịch gian lận đã được gán nhãn → Classification (XGBoost).</p>
+</blockquote>
+
+<h2 id="recommendation"><strong>3. Recommendation Systems</strong></h2>
+
+<table>
+<thead><tr><th>Type</th><th>Algorithm</th><th>AWS Service</th></tr></thead>
+<tbody>
+<tr><td>Collaborative Filtering</td><td>Factorization Machines, Neural CF</td><td>SageMaker FM, <strong>Amazon Personalize</strong></td></tr>
+<tr><td>Content-Based</td><td>TF-IDF, Embeddings</td><td>SageMaker Knn</td></tr>
+<tr><td>Hybrid</td><td>FM + content features</td><td>Amazon Personalize (HRNN)</td></tr>
+<tr><td>Cold Start Problem</td><td>Content features, metadata</td><td>Personalize context metadata</td></tr>
+</tbody>
+</table>
+
+<p>Đề thi thường hỏi: <em>"Real-time personalized recommendations for e-commerce"</em> → câu trả lời gần như luôn là <strong>Amazon Personalize</strong> (managed service).</p>
+
+<h2 id="nlp"><strong>4. NLP Pipeline</strong></h2>
+
+<pre><code class="language-text">NLP Task Decision Tree:
+
+Classify documents?
+    → Text Classification → BlazingText (word2vec mode), Comprehend Custom
+
+Extract entities from text?
+    → Named Entity Recognition → Amazon Comprehend (NER)
+
+Translate text?
+    → Amazon Translate
+
+Summarize documents?
+    → Hugging Face on SageMaker (T5, BART)
+
+Q&A / Generation?
+    → SageMaker JumpStart foundation models (Llama, Falcon)
+
+Toxic content detection?
+    → Amazon Comprehend (Sentiment + custom classifier)
+</code></pre>
+
+<table>
+<thead><tr><th>NLP Task</th><th>Algorithm/Service</th></tr></thead>
+<tbody>
+<tr><td>Document classification</td><td>BlazingText, Comprehend Custom</td></tr>
+<tr><td>Topic modeling</td><td><strong>Latent Dirichlet Allocation (LDA)</strong>, NTM</td></tr>
+<tr><td>Sentiment analysis</td><td>Amazon Comprehend</td></tr>
+<tr><td>Language detection</td><td>Amazon Comprehend</td></tr>
+<tr><td>Machine translation</td><td>Amazon Translate</td></tr>
+<tr><td>Conversational AI</td><td>Amazon Lex (chatbot) + Lambda</td></tr>
+</tbody>
+</table>
+
+<h2 id="time-series"><strong>5. Time Series Forecasting</strong></h2>
+
+<table>
+<thead><tr><th>Service/Algorithm</th><th>Use Case</th></tr></thead>
+<tbody>
+<tr><td><strong>DeepAR+</strong> (SageMaker built-in)</td><td>Multiple related time series, cold start</td></tr>
+<tr><td><strong>Amazon Forecast</strong></td><td>Fully managed, business forecasting (demand, inventory)</td></tr>
+<tr><td><strong>ARIMA</strong></td><td>Single stationary series, classic approach</td></tr>
+<tr><td><strong>Prophet</strong></td><td>Seasonality + holidays, Facebook's library</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> "Predict retail demand for thousands of products" → <strong>DeepAR+</strong> (multiple related series) or <strong>Amazon Forecast</strong> (managed). Key differentiator: DeepAR+ trains one model for ALL items, while classical methods need one model per series.</p>
+</blockquote>
+
+<h2 id="computer-vision"><strong>6. Computer Vision</strong></h2>
+
+<table>
+<thead><tr><th>CV Task</th><th>Algorithm</th><th>Service</th></tr></thead>
+<tbody>
+<tr><td>Image Classification</td><td><strong>Image Classification (ResNet)</strong></td><td>SageMaker built-in</td></tr>
+<tr><td>Object Detection</td><td><strong>Object Detection (SSD/YOLO)</strong></td><td>SageMaker built-in</td></tr>
+<tr><td>Semantic Segmentation</td><td><strong>Semantic Segmentation</strong></td><td>SageMaker built-in</td></tr>
+<tr><td>Rekognition (no ML needed)</td><td>Faces, objects, text, moderation</td><td><strong>Amazon Rekognition</strong></td></tr>
+<tr><td>Custom labels in images</td><td>Fine-tuning</td><td>Rekognition Custom Labels</td></tr>
+</tbody>
+</table>
+
+<h2 id="mapping-table"><strong>7. Business Domain → AWS Service Mapping</strong></h2>
+
+<table>
+<thead><tr><th>Industry / Scenario</th><th>AWS Service</th></tr></thead>
+<tbody>
+<tr><td>E-commerce personalization</td><td>Amazon Personalize</td></tr>
+<tr><td>Call center analytics</td><td>Amazon Transcribe + Comprehend</td></tr>
+<tr><td>Medical image analysis</td><td>Amazon HealthLake + SageMaker (custom model)</td></tr>
+<tr><td>Document understanding (invoices, forms)</td><td><strong>Amazon Textract</strong></td></tr>
+<tr><td>Product search (semantic)</td><td>Amazon OpenSearch + KNN</td></tr>
+<tr><td>IoT anomaly detection</td><td><strong>Amazon Lookout for Equipment</strong></td></tr>
+<tr><td>Manufacturing defect detection</td><td><strong>Amazon Lookout for Vision</strong></td></tr>
+<tr><td>Financial forecasting (managed)</td><td>Amazon Forecast</td></tr>
+<tr><td>Churn prediction</td><td>XGBoost (SageMaker)</td></tr>
+</tbody>
+</table>
+
+<h2 id="practice"><strong>8. Practice Questions</strong></h2>
+
+<p><strong>Q1:</strong> A retail company wants to predict next quarter's sales for 10,000 individual products, incorporating holiday seasonality and promotional events. Which solution is MOST appropriate?</p>
+<ul>
+<li>A) Train separate ARIMA models for each of the 10,000 products</li>
+<li>B) Use Amazon Forecast with related time series data ✓</li>
+<li>C) Use Amazon Comprehend for trend analysis</li>
+<li>D) Use Amazon Rekognition to analyze product images</li>
+</ul>
+<p><em>Explanation: Amazon Forecast is designed for business demand forecasting at scale, can handle thousands of related time series simultaneously, and supports external factors like holidays and promotions. Training 10,000 separate ARIMA models would be impractical and miss cross-series patterns.</em></p>
+
+<p><strong>Q2:</strong> A bank needs to detect fraudulent transactions in real time WITHOUT having labeled historical fraud data. Which algorithm should they use?</p>
+<ul>
+<li>A) XGBoost classifier</li>
+<li>B) Linear Learner with binary classification</li>
+<li>C) Random Cut Forest ✓</li>
+<li>D) BlazingText</li>
+</ul>
+<p><em>Explanation: Without labeled fraud data, this is an unsupervised anomaly detection problem. Random Cut Forest detects anomalies (transactions that deviate from normal patterns) without requiring labeled examples. XGBoost and Linear Learner are supervised and require labeled training data.</em></p>
+
+<p><strong>Q3:</strong> A company wants to extract key-value pairs from scanned insurance claim forms (PDFs). No custom ML training has been done. Which AWS service should they use?</p>
+<ul>
+<li>A) Amazon Comprehend</li>
+<li>B) Amazon Textract ✓</li>
+<li>C) SageMaker Object Detection</li>
+<li>D) Amazon Rekognition</li>
+</ul>
+<p><em>Explanation: Amazon Textract is purpose-built for extracting text, forms (key-value pairs), and tables from scanned documents including PDFs. It requires no ML training. Comprehend analyzes text meaning; Rekognition analyzes images; neither handles structured form extraction like Textract.</em></p>

package/content/series/luyen-thi/luyen-thi-aws-ml-specialty/chapters/04-phan-4-on-tap/lessons/11-bai-11-cheat-sheet.md

@@ -0,0 +1,110 @@
+---
+id: 722555b5-e196-45c3-9061-a4f72197e2ed
+title: 'Bài 11: Cheat Sheet Tổng Hợp'
+slug: bai-11-cheat-sheet
+description: >-
+  Bảng tổng hợp toàn bộ khoá học: thuật toán SageMaker, AWS AI services,
+  evaluation metrics, công thức quan trọng và các bẫy thường gặp trong đề thi.
+duration_minutes: 40
+is_free: true
+video_url: null
+sort_order: 11
+section_title: "Phần 4: Ôn Tập Tổng Hợp"
+course:
+  id: 019c9619-lt02-7002-c002-lt0200000002
+  title: 'Luyện thi AWS Certified Machine Learning - Specialty'
+  slug: luyen-thi-aws-ml-specialty
+---
+
+<h2 id="algorithms-master"><strong>1. SageMaker Built-in Algorithms Master Table</strong></h2>
+
+<table>
+<thead><tr><th>Algorithm</th><th>Type</th><th>Best For</th><th>Input</th></tr></thead>
+<tbody>
+<tr><td><strong>XGBoost</strong></td><td>Supervised (C/R)</td><td>Tabular data, competitions</td><td>CSV, LibSVM</td></tr>
+<tr><td><strong>Linear Learner</strong></td><td>Supervised (C/R)</td><td>High-dimensional, fast</td><td>CSV, RecordIO</td></tr>
+<tr><td><strong>Random Cut Forest</strong></td><td>Unsupervised</td><td>Anomaly detection</td><td>CSV, RecordIO</td></tr>
+<tr><td><strong>K-Means</strong></td><td>Unsupervised</td><td>Customer segmentation</td><td>CSV, RecordIO</td></tr>
+<tr><td><strong>PCA</strong></td><td>Dimensionality Reduction</td><td>Feature reduction</td><td>CSV, RecordIO</td></tr>
+<tr><td><strong>Factorization Machines</strong></td><td>Supervised (C/R)</td><td>Sparse data, recommendations</td><td>RecordIO only</td></tr>
+<tr><td><strong>DeepAR+</strong></td><td>Supervised</td><td>Time series forecasting</td><td>JSON Lines</td></tr>
+<tr><td><strong>BlazingText</strong></td><td>Supervised / Unsupervised</td><td>Text classification, word2vec</td><td>Text files</td></tr>
+<tr><td><strong>Object2Vec</strong></td><td>Supervised</td><td>Semantic similarity</td><td>JSON Lines</td></tr>
+<tr><td><strong>Image Classification</strong></td><td>Supervised</td><td>Image labels</td><td>RecordIO, raw images</td></tr>
+<tr><td><strong>Object Detection</strong></td><td>Supervised</td><td>Bounding boxes</td><td>RecordIO, JSON</td></tr>
+<tr><td><strong>Semantic Segmentation</strong></td><td>Supervised</td><td>Pixel-level classification</td><td>Images + masks</td></tr>
+<tr><td><strong>LDA</strong></td><td>Unsupervised</td><td>Topic modeling</td><td>CSV, RecordIO</td></tr>
+<tr><td><strong>NTM</strong></td><td>Unsupervised</td><td>Neural topic modeling</td><td>CSV, RecordIO</td></tr>
+<tr><td><strong>IP Insights</strong></td><td>Unsupervised</td><td>IP address anomaly</td><td>CSV</td></tr>
+</tbody>
+</table>
+
+<h2 id="aws-ai-services"><strong>2. AWS AI Services — No-Code ML</strong></h2>
+
+<table>
+<thead><tr><th>Service</th><th>Purpose</th><th>Output</th></tr></thead>
+<tbody>
+<tr><td><strong>Amazon Rekognition</strong></td><td>Image/video analysis</td><td>Labels, faces, text, moderation</td></tr>
+<tr><td><strong>Amazon Textract</strong></td><td>Document extraction</td><td>Text, forms, tables</td></tr>
+<tr><td><strong>Amazon Comprehend</strong></td><td>NLP, text analytics</td><td>Entities, sentiment, topics</td></tr>
+<tr><td><strong>Amazon Translate</strong></td><td>Machine translation</td><td>Translated text</td></tr>
+<tr><td><strong>Amazon Transcribe</strong></td><td>Speech to text</td><td>Transcription, subtitles</td></tr>
+<tr><td><strong>Amazon Polly</strong></td><td>Text to speech</td><td>Audio</td></tr>
+<tr><td><strong>Amazon Lex</strong></td><td>Conversational AI (chatbot)</td><td>Intent, slots</td></tr>
+<tr><td><strong>Amazon Kendra</strong></td><td>Intelligent search</td><td>Answers, documents</td></tr>
+<tr><td><strong>Amazon Personalize</strong></td><td>Recommendations</td><td>Item rankings, user recs</td></tr>
+<tr><td><strong>Amazon Forecast</strong></td><td>Time series forecasting</td><td>Predictions + confidence</td></tr>
+<tr><td><strong>Amazon Lookout for Vision</strong></td><td>Visual anomaly (manufacturing)</td><td>Pass/Fail, anomaly map</td></tr>
+<tr><td><strong>Amazon Lookout for Equipment</strong></td><td>Equipment anomaly (IoT)</td><td>Anomaly scores</td></tr>
+</tbody>
+</table>
+
+<h2 id="metrics"><strong>3. Evaluation Metrics Quick Reference</strong></h2>
+
+<table>
+<thead><tr><th>Metric</th><th>Formula</th><th>Use When</th></tr></thead>
+<tbody>
+<tr><td><strong>Accuracy</strong></td><td>(TP+TN)/(TP+TN+FP+FN)</td><td>Balanced classes</td></tr>
+<tr><td><strong>Precision</strong></td><td>TP/(TP+FP)</td><td>FP is costly (spam filter)</td></tr>
+<tr><td><strong>Recall (Sensitivity)</strong></td><td>TP/(TP+FN)</td><td>FN is costly (cancer diagnosis)</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 TPR vs FPR curve</td><td>Overall classifier quality</td></tr>
+<tr><td><strong>RMSE</strong></td><td>√(Σ(yᵢ-ŷᵢ)²/n)</td><td>Regression, penalizes outliers</td></tr>
+<tr><td><strong>MAE</strong></td><td>Σ|yᵢ-ŷᵢ|/n</td><td>Regression, robust to outliers</td></tr>
+<tr><td><strong>MAPE</strong></td><td>Σ|yᵢ-ŷᵢ|/|yᵢ| × 100%</td><td>Forecasting, interpretable %</td></tr>
+</tbody>
+</table>
+
+<h2 id="traps"><strong>4. Common Exam Traps</strong></h2>
+
+<table>
+<thead><tr><th>Trap</strong></th><th>What the Exam Says</th><th>Correct Answer</th></tr></thead>
+<tbody>
+<tr><td>Imbalanced data + accuracy</td><td>"Model has 99% accuracy" (fraud)</td><td>Use Precision/Recall/F1, not accuracy</td></tr>
+<tr><td>FM input format</td><td>Factorization Machines</td><td>Requires RecordIO ONLY (not CSV)</td></tr>
+<tr><td>Managed vs custom</td><td>"quickest to implement"</td><td>Prefer managed (Personalize, Forecast)</td></tr>
+<tr><td>Overfitting fix</td><td>Training accuracy high, validation low</td><td>Regularization (L1/L2) or more data</td></tr>
+<tr><td>SageMaker + Internet</td><td>"secure environment, no internet"</td><td>VPC + Network Isolation + VPC Endpoints</td></tr>
+<tr><td>Ground Truth labeling</td><td>"reduce labeling cost"</td><td>Auto-labeling (active learning) in GT</td></tr>
+<tr><td>Bias in model</td><td>"identify bias before deployment"</td><td>SageMaker Clarify</td></tr>
+<tr><td>Multiple models behind one endpoint</td><td>"save cost, single endpoint"</td><td>Multi-Model Endpoint (MME)</td></tr>
+</tbody>
+</table>
+
+<h2 id="storage-formats"><strong>5. Data / Storage Quick Reference</strong></h2>
+
+<table>
+<thead><tr><th>Scenario</th><th>Best Choice</th></tr></thead>
+<tbody>
+<tr><td>Large tabular training data</td><td>S3 + CSV or Parquet; RecordIO for SageMaker</td></tr>
+<tr><td>Real-time streaming data ingestion</td><td>Kinesis Data Streams → Firehose → S3</td></tr>
+<tr><td>Ad-hoc SQL queries on S3 data</td><td>Amazon Athena</td></tr>
+<tr><td>ETL transformation → Feature Store</td><td>AWS Glue (Spark ETL) → SageMaker Feature Store</td></tr>
+<tr><td>Business intelligence / dashboards</td><td>Amazon QuickSight</td></tr>
+<tr><td>Data warehouse for ML features</td><td>Amazon Redshift → ML (Redshift ML)</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> Hãy nhớ: khi đề bài nói "fastest / easiest / no code" → AWS Managed AI Service. Khi nói "custom model / flexibility / bring your own" → SageMaker. Đây là decision boundary quan trọng nhất.</p>
+</blockquote>

package/content/series/luyen-thi/luyen-thi-aws-ml-specialty/chapters/04-phan-4-on-tap/lessons/12-bai-12-chien-luoc-thi.md

@@ -0,0 +1,113 @@
+---
+id: 87d7c5c5-d20a-47a8-b64e-7f9db75bb1d2
+title: 'Bài 12: Chiến Lược Thi MLS-C01'
+slug: bai-12-chien-luoc-thi
+description: >-
+  Cấu trúc đề thi MLS-C01, phân bổ điểm số theo domain, chiến lược quản lý
+  thời gian, kỹ thuật loại trừ đáp án và kế hoạch ôn tập cuối kỳ.
+duration_minutes: 35
+is_free: true
+video_url: null
+sort_order: 12
+section_title: "Phần 4: Ôn Tập Tổng Hợp"
+course:
+  id: 019c9619-lt02-7002-c002-lt0200000002
+  title: 'Luyện thi AWS Certified Machine Learning - Specialty'
+  slug: luyen-thi-aws-ml-specialty
+---
+
+<h2 id="exam-structure"><strong>1. Cấu Trúc Đề Thi MLS-C01</strong></h2>
+
+<table>
+<thead><tr><th>Item</th><th>Details</th></tr></thead>
+<tbody>
+<tr><td><strong>Total Questions</strong></td><td>65 câu</td></tr>
+<tr><td><strong>Scored Questions</strong></td><td>50 câu (15 unscored/experimental)</td></tr>
+<tr><td><strong>Time Limit</strong></td><td>180 phút (3 giờ)</td></tr>
+<tr><td><strong>Passing Score</strong></td><td>750/1000</td></tr>
+<tr><td><strong>Question Format</strong></td><td>Multiple choice + Multiple response</td></tr>
+<tr><td><strong>Languages</strong></td><td>Có thể chọn tiếng Nhật, tiếng Hàn (không có tiếng Việt)</td></tr>
+<tr><td><strong>Validity</strong></td><td>3 năm</td></tr>
+</tbody>
+</table>
+
+<h2 id="domain-weights"><strong>2. Domain Weights — Phân Bổ Điểm</strong></h2>
+
+<table>
+<thead><tr><th>Domain</th><th>Weight</th><th>Questions (~)</th></tr></thead>
+<tbody>
+<tr><td>Domain 1: Data Engineering</td><td>20%</td><td>~13 câu</td></tr>
+<tr><td>Domain 2: Exploratory Data Analysis</td><td>24%</td><td>~16 câu</td></tr>
+<tr><td>Domain 3: Modeling</td><td>36%</td><td>~24 câu</td></tr>
+<tr><td>Domain 4: ML Implementation & Operations</td><td>20%</td><td>~13 câu</td></tr>
+</tbody>
+</table>
+
+<blockquote>
+<p><strong>Exam tip:</strong> <strong>Domain 3 (Modeling) chiếm 36%</strong> — đây là domain quan trọng nhất. Ưu tiên ôn: SageMaker built-in algorithms, evaluation metrics, hyperparameter tuning, và underfitting/overfitting.</p>
+</blockquote>
+
+<h2 id="time-management"><strong>3. Chiến Lược Quản Lý Thời Gian</strong></h2>
+
+<pre><code class="language-text">Time Budget (180 minutes, 65 questions):
+
+Average: 2.5 minutes per question
+
+Strategy:
+Round 1 (120 min): Attempt all 65 questions
+  - Answer confident questions immediately
+  - Flag uncertain questions (don't spend > 3 min each)
+  
+Round 2 (45 min): Review flagged questions
+  - Apply elimination technique
+  - Check scenario keywords
+
+Round 3 (15 min): Final review
+  - Check unmarked questions
+  - Don't change answers unless certain
+</code></pre>
+
+<h2 id="elimination"><strong>4. Kỹ Thuật Loại Trừ Đáp Án</strong></h2>
+
+<table>
+<thead><tr><th>Keyword in Question</th><th>Eliminate</th><th>Keep / Prefer</th></tr></thead>
+<tbody>
+<tr><td>"no labels / unlabeled data"</td><td>Supervised algorithms</td><td>K-Means, RCF, PCA</td></tr>
+<tr><td>"quickest / easiest / no code"</td><td>Custom SageMaker models</td><td>Managed AI services</td></tr>
+<tr><td>"real-time predictions"</td><td>Batch Transform</td><td>Real-time Endpoint</td></tr>
+<tr><td>"low traffic, cost-efficient inference"</td><td>Real-time Endpoint</td><td>Serverless Inference</td></tr>
+<tr><td>"no internet, secure"</td><td>Public endpoint</td><td>VPC + Network Isolation</td></tr>
+<tr><td>"imbalanced dataset"</td><td>Accuracy metric</td><td>F1, Precision, Recall, AUC</td></tr>
+<tr><td>"multiple models, save cost"</td><td>Separate endpoints</td><td>Multi-Model Endpoint</td></tr>
+<tr><td>"structured tabular data"</td><td>Image/NLP algorithms</td><td>XGBoost, Linear Learner</td></tr>
+</tbody>
+</table>
+
+<h2 id="study-plan"><strong>5. Kế Hoạch Ôn Tập Cuối Kỳ</strong></h2>
+
+<table>
+<thead><tr><th>Ngày</th><th>Focus</th></tr></thead>
+<tbody>
+<tr><td>Day 1</td><td>Review Domain 3 (Modeling): algorithms + metrics + HPO</td></tr>
+<tr><td>Day 2</td><td>Review Domain 2 (EDA): data preparation + feature engineering</td></tr>
+<tr><td>Day 3</td><td>Review Domain 1+4 (Data Engineering + Operations)</td></tr>
+<tr><td>Day 4</td><td>Practice exam 1 (65 câu): identify weak areas</td></tr>
+<tr><td>Day 5</td><td>Review weak areas từ Practice Exam 1</td></tr>
+<tr><td>Day 6</td><td>Practice exam 2 (65 câu)</td></tr>
+<tr><td>Day 7</td><td>Cheat sheet review only — nghỉ ngơi, không học mới</td></tr>
+</tbody>
+</table>
+
+<h2 id="day-of-exam"><strong>6. Ngày Thi</strong></h2>
+
+<ul>
+<li>Đến sớm 30 phút (Pearson VUE test center) hoặc check kỹ environment (online proctored)</li>
+<li>Đọc kỹ toàn bộ câu hỏi — đừng chỉ đọc 2-3 dòng đầu</li>
+<li>Highlight keywords: "MOST cost-effective", "LEAST operational overhead", "FASTEST to deploy"</li>
+<li>Khi thấy 2 đáp án đều đúng: chọn cái phù hợp nhất với constraints trong câu hỏi</li>
+<li>Không để trống câu nào — guess nếu không biết (không bị trừ điểm)</li>
+</ul>
+
+<blockquote>
+<p><strong>Exam tip:</strong> AWS Specialty exams thường có câu hỏi dài (3-5 dòng scenario). Keywords ở CUỐI câu thường là key constraint: "with minimal operational overhead", "in the most cost-effective manner", "without requiring data scientists" — đọc kỹ đến cuối!</p>
+</blockquote>

package/content/series/luyen-thi/luyen-thi-aws-ml-specialty/index.md

@@ -7,7 +7,7 @@ description: >-
   (MLS-C01). Nắm vững SageMaker, data engineering, modeling, triển khai ML trên AWS
   ở mức chuyên gia.
 
-featured_image: null
+featured_image: images/blog/aws-ml-specialty-series-banner.png
 level: advanced
 duration_hours: 40
 lesson_count: 12